Methods

Methods compendium

How every Data Garden value is measured. Each method opens with its principle, instrument and what it reports; the full principle, procedure, equipment and references fold out below it. The text is the compendium's, version 1.20 (2026-10-29).

Abstract

Here we present the methods compendium for the Data Garden, an open analytical platform that links plant protein sources to aqueous extraction methods and to functional outcomes in food-relevant assays. The platform applies a consistent protocol set - three aqueous extraction routes and a battery of compositional, structural, and functional assays - to a growing library of plant-protein materials, prioritizing reliability and comparability over throughput. All extracts are analyzed in the wet state, and composition and assay loadings are reported on that basis. Results in the Data Garden trace back to methods outlined in this document. This document records the canonical method definitions, the data reported for each method, the procedural steps taken, and supporting literature.

In the PDF, p. 2

Introduction

Industrial protein extraction is typically optimized for yield rather than functionality. Protein isolates that succeed in yield optimization frequently fail in the functional properties that food applications need most - stretch and melt for cheese analogues, binding and foaming for egg analogues, solubility and emulsion stability across dairy bases and beverages. What has been missing is a shared, openly reported evidence base that links extraction parameters back to those functional outcomes across many crops under one protocol set. The Data Garden platform is built to close that gap.

Read the rest of this section

By focusing the core extraction workflow on three aqueous routes - isoelectric precipitation, micellar protein isolation, and membrane-based concentration - we aim to reveal the range of functional diversity within a single protein source and across commodity-level plant protein sources in the Data Garden catalog. Analytical readouts are reported with the assay conditions that produced them (pH, dispersion medium, protein loading, instrument method) so any downstream reader can interpret a result against those conditions.

Our replication strategy prioritizes coverage across sources and routes over replication within a single source. A given protein source is extracted once per route, and replicates reported in our data refer universally to technical replicates. Where possible we measure multiple examples of the same protein source, such as pea flour from several suppliers, which tests whether a commodity-level signal survives a change of material even though it does not replicate the extraction itself. This work operates on a screening paradigm: it seeks signals that imply improved functionality and identifies conditions worth pursuing. The first dataset starts from commercial flours and two commercial protein concentrates, each put through all three extraction routes. Commercial isolates follow in a second dataset; their protein is already separated, so they are not extracted but hydrated and tested as supplied (see Hydration of Commercial Isolates). Flours are less sensitive to a manufacturer's processing history than isolates are, but the crop itself still varies with cultivar, growing season, and storage, so values for a commodity are reported as the range observed across the materials tested and are not claimed to represent every material sold under that name.

Unless a method states otherwise, loadings expressed as percent protein (w/w) mean grams of protein per 100 g of dispersion, with protein content determined by combustion (see Nitrogen Concentration). Normalizing to protein rather than to sample mass compares materials of differing purity at equal protein, which matters most for interfacial and gelling measurements. For methods where that normalization is not essential, the next screening round moves to a sample basis (w/v), which is simpler to reproduce in other laboratories.

Our scientific advisors have proposed measurements beyond the current protocol set. A viscosity assay for liquid applications will be added in the next screening round. Confocal microscopy of gel and emulsion microstructure would be valuable but is not currently available to us. Small-deformation rheology is not planned; gel texture is characterized by texture profile analysis (see Gel Strength). Where other laboratories hold the instrumentation, measurements run on the same materials under a standardized protocol would be a welcome complement to the dataset.

Each method in this compendium includes a brief statement on the principle and rationale, the data types recorded or generated, the equipment and consumables required, an operational outline of the bench-level procedure, and supporting references. Each method description in this compendium corresponds to a versioned method template that has been reviewed and approved by the Data Garden team and stored in LIMBBS; template identifiers and versions are available on request. Expect continued iteration as our methods mature; we very much welcome your thoughts. Critical review only makes the work better - please direct correspondence to info@sustainableprotein.org

In the PDF, p. 3

Aqueous Protein Extraction

Seed storage proteins are the dominant dietary protein fraction in nearly every seed-derived crop humans eat. They accumulate during seed development in specialized vacuoles, where they remain inert until germination triggers proteolytic mobilization. Three structural classes dominate the legume and oilseed isolates that anchor the Data Garden: the 11S legumin-type globulins, the 7S vicilin-type globulins, and the 2S albumins. The balance among them is strongly crop-dependent, with the 7S and 11S globulins carrying most of the storage protein in legumes and the 2S albumins more prominent in several oilseeds. Their surface charge, hydrophobic patch distribution, and quaternary-structure assemblies at neutral pH set the floor for what wet-process isolation can recover and shape every downstream functional behavior.

Read the rest of this section

Three aqueous routes are currently in use across the Data Garden: Isoelectric Precipitation (IPI), Micellar Protein Isolation (MPI), and Membrane-based Concentration (MEM). Each starts from a plant flour, defatted where the commodity requires it, or from a commercial protein concentrate. Concentrates are charged exactly as flours are, at 10% (w/w) powder in a 2000 g solubilization (200 g of powder per route), so a batch is matched on powder mass, not on protein. A concentrate therefore brings more protein into each batch than a flour does, and yields are calculated against the protein actually charged (see Mass Balance and Yield Accounting). The chemistry of each route shapes the functional behavior of the product in distinct and often non-trivial ways. The same starting material can yield isolates with very different solubility curves, gel-strength ceilings, and emulsification capacities depending on which route was used - a fact that justifies treating extraction route as a primary variable in the dataset.

The sub-sections that follow describe each route, its underlying chemistry, and its known strengths and weaknesses for downstream applications. The bench-level procedure - solubilization, separation, concentration, and wash - is described for each route.

All three routes create a wet extract: a washed pellet for IPI and MPI, a liquid retentate for MEM. Extracts are not neutralized and are not dried. They are stored at 4 °C and carried into compositional and functional assays in the wet state, with protein content determined on the wet material by combustion (see Nitrogen Concentration) and assay loadings standardized on that basis. This is a deliberate choice, since drying is itself a processing step that alters solubility and structure, and omitting it keeps the extraction route as the dominant variable distinguishing one extract from another. It also means the extract pH at collection is route-dependent, near the precipitation pH for IPI and near the solubilization pH for MEM, and is recorded for every extract rather than adjusted to a common value. Functional assays that require a defined pH adjust the dispersion at the point of assay, and that adjustment is recorded with the result.

References

  1. McClements, D. J., & Grossmann, L. (2022). Next-Generation Plant-based Foods: Design, Production, and Properties. Springer. https://doi.org/10.1007/978-3-030-96764-2
  2. Kimura, A., Fukuda, T., Zhang, M., Motoyama, S., Maruyama, N., & Utsumi, S. (2008). Comparison of physicochemical properties of 7S and 11S globulins from pea, fava bean, cowpea, and French bean with those of soybean: French bean 7S globulin exhibits excellent properties. Journal of Agricultural and Food Chemistry, 56(21), 10273–10279. https://doi.org/10.1021/jf801721b
  3. Souza, P. F. N. (2020). The forgotten 2S albumin proteins: importance, structure, and biotechnological application in agriculture and human health. International Journal of Biological Macromolecules, 164, 4638–4649. https://doi.org/10.1016/j.ijbiomac.2020.09.049
  4. Yang, J., Zamani, S., Liang, L., & Chen, L. (2021). Extraction methods significantly impact pea protein composition, structure and gelling properties. Food Hydrocolloids, 117, 106678. https://doi.org/10.1016/j.foodhyd.2021.106678
  5. Stone, A. K., Karalash, A., Tyler, R. T., Warkentin, T. D., & Nickerson, M. T. (2015). Functional attributes of pea protein isolates prepared using different extraction methods and cultivars. Food Research International, 76, 31–38. https://doi.org/10.1016/j.foodres.2014.11.017

Yield Screen

Extraction conditions determine how much protein is recovered from a given starting material, and the optimum differs by commodity. This method screens those conditions at small scale before committing to a full extraction, using a sweep of solubilization pH against NaCl concentration. Precipitation pH is not screened, rather proteins are precipitated at the literature-determined isoelectric pH for a given source, which is held constant across the sweep and carried forward unchanged to full-scale extraction.

Instrument

LECO FP828 Nitrogen Analyzer

What is reported

  • The salt and pH levels used for this commodity
  • Protein yield (%) for each of the conditions screened
  • Best Yield Conditions: Solubilization pH and Salt Concentration (M)
  • Precipitation pH used for the salt-free arm, and its source
  • Measured pH of the unadjusted condition at each salt level
Full principle

The levels given below are a default starting grid. Where literature for a commodity indicates that its optimum lies outside that range, the levels may be adjusted before the screen is run, and the screening conditions are recorded with the result.

Protein solubility in seed flours depends on both pH and ionic strength. Raising pH above the isoelectric point increases net charge and electrostatic repulsion, driving globulins into solution. Adding salt has a more complex effect: at low concentrations it screens charge and increases solubility, the salting-in regime that defines globulins as a class, while at higher concentrations it competes for water and reduces solubility. The two variables interact, so the useful optimum is found empirically rather than predicted.

The two salt regimes also dictate how protein is recovered. Salt-free extracts are precipitated isoelectrically, by lowering pH to the point of minimum net charge. Salt extracts are recovered instead by micellization, since dissolved salt substantially suppresses isoelectric precipitation: diluting the extract into cold water drops the ionic strength abruptly, which drives the protein to self-associate into micelle-like assemblies that sediment on centrifugation. Recovery route and salt level are therefore linked by this design rather than varied independently, and the screen compares whole extraction routes rather than isolating the effect of salt alone. A comparison between a salt-free and a salted cell reflects both the change in salt and the change in recovery route.

Nitrogen is determined on both the starting ingredient and the recovered pellet, and yield is the fraction of the starting protein recovered. Conditions giving the highest yield are carried forward as the starting point for full-scale extraction. Because the screen is run without replication and at a scale where handling losses are proportionally larger, it identifies a promising region rather than a precise optimum, and the selected condition is confirmed at scale.

The screen is an abbreviated form of the full routes rather than a miniature of them. The micellization arm omits the overnight settling used at full scale, both arms omit the washes, and centrifugation is at a lower field than the extraction protocols use. Pellets are unwashed, so retained liquid and salt contribute to the recovered mass and a wetter or saltier pellet can appear to recover more. Recovered nitrogen also includes co-recovered non-protein nitrogen. The screen therefore ranks conditions rather than predicting the yield the full protocol will deliver.

Protein yield (%) = (pellet mass × pellet protein content) / (starting powder mass × starting powder protein content) × 100, with both protein contents determined by combustion on the material as analyzed (see Nitrogen Concentration).

The selected condition is the highest-yielding cell within the recovery route the material is destined for, not the highest cell overall, since the salt-free and salted cells feed different routes and are not a single comparable response surface.

Equipment
  • LECO FP828 Nitrogen Analyzer
  • Analytical balance
  • Water bath
  • Centrifuge (Thermo Scientific TX-400 Rotor)
  • pH meter (Hanna Instruments Foodcare; see Appendix B)
Consumables
  • 15 mL centrifuge tubes
  • 50 mL centrifuge tubes
  • NaCl (Diamond Crystal Kosher Salt, ≥99% NaCl; molarity calculated on that basis)
  • 1 M NaOH/ 1 M HCl
Procedure
  1. Sample preparation. Prepare four salt solutions. The default levels are 0.0, 0.2, 0.5, and 1.0 M NaCl. Label a set of 16, 15 mL centrifuge tubes. Add 10% (w/w) powder to each salt solution for 12 g total. Make 4 tubes per salt concentration for the pH sweep. Record initial powder mass. Adjust pH across four levels. The defaults are 7.0, 8.0, and 9.0, plus one unadjusted tube per salt level, recording the measured pH of the unadjusted tubes. Heat samples in a 40 °C water bath for 1 hour. Check pH, and readjust as needed.
  2. Tube weight. Label another set of 16, 15 mL tubes, and weigh. Label a set of 12, 50 mL tubes for micellization, and weigh.
  3. First spin - remove insolubles. Centrifuge samples for 25 min at 1,690 × g (3000 rpm, TX-400 rotor) and 20 °C. Immediately transfer the solubilized protein supernatant into the new set of 15 mL tubes.
  4. Isoelectric precipitation. For the 4 no-salt samples, drop pH to the source’s literature isoelectric precipitation pH with 1M HCl. Record the value used and its source.
  5. Micellar precipitation. For the 12 salt samples, dilute supernatant in 5x cold DI water (4 °C) using the set of 50 mL tubes. Transfer no more than 8 mL of supernatant per tube so that a full fivefold dilution fits within the 50 mL volume.
  6. Second spin. Centrifuge all samples for 25 min at 1,690 × g (3000 rpm, TX-400 rotor) and 20 °C. Immediately pour off the supernatant and discard. Weigh the final pellet plus tube mass.
  7. LECO. Test the pellet for protein content on the LECO to calculate the protein yield of each condition.
References
  1. Karaca, A. C., Low, N., & Nickerson, M. (2011). Emulsifying properties of chickpea, faba bean, lentil and pea proteins produced by isoelectric precipitation and salt extraction. Food Research International 44(9), 2742-2750.
  2. Stone, A. K., Karalash, A., Tyler, R. T., Warkentin, T. D., & Nickerson, M. T. (2015). Functional attributes of pea protein isolates prepared using different extraction methods and cultivars. Food Research International 76, 31-38.
  3. Boye, J., Zare, F., & Pletch, A. (2010). Pulse proteins: processing, characterization, functional properties and applications in food and feed. Food Research International 43(2), 414-431.
  4. Murray, E. D., Myers, C. D., Barker, L. D., & Maurice, T. J. (1980). Process for isolation of proteins using food grade salt solutions at specified pH and ionic strength. US Patent 4,208,323.
In the PDF, p. 6

Isoelectric Precipitation

Isoelectric precipitation exploits the pH dependence of a protein's net charge. When the pH of a protein extract is alkaline, the globulin proteins carry a net negative charge as their carboxylate residues (Asp, Glu, C terminus) are deprotonated and outnumber protonated basic groups (Lys, Arg, His). Acidifying the solution towards the isoelectric point (pI) causes more of these carboxylates to become protonated, reducing the negative charge. At the pI, usually between pH 4-5.5 for seed globulins, negative and positive contributions are balanced, and the zeta potential of the protein approaches zero.

What is reported

  • Solubilization conditions: Initial pH, Target pH, Final pH
  • Precipitation conditions: Initial pH, Target pH, Final pH
  • Solubilized & sieved extract mass (g)
  • Solubilized & sieved extract protein content (%)
  • Extract pH
  • Extract observations: Color, texture
  • Solubilization efficiency (%)
  • Number of washes
  • Final supernatant conductivity (mS/cm)
  • Protein yield

Protein content and protein yield for extracts produced by this route are determined as described under Nitrogen Concentration (Dumas Method). Routine pH, conductivity, and mass measurements follow Appendix B.

Full principle

At this point, electrostatic repulsion no longer offsets hydrophobic and hydrogen bonding interactions and solubility passes through a minimum. The protein aggregates and falls out of solution. The precipitate is then recovered by centrifugation and washed. Recovery is generally high, but incomplete, since albumins and low molecular weight peptides remain soluble at globulin pI and are lost to the supernatant.

Strengths. Simple, scalable, and reliable. Only inexpensive reagents (NaOH and HCl), although the dominant process costs are water, dewatering, and drying rather than chemicals. Recovery from the alkaline extract into the isolate is typically 70-85%, corresponding to roughly 55-70% of the available protein in the starting flour. The product is a high-purity isolate, generally above 90% protein for soy and 80-90% for most pulses, on a dry basis*.

Weaknesses. The isolate is structurally altered rather than native. Alkaline extraction dissociates the oligomeric globulins and acid-induced aggregation is only partly reversible, so non-native aggregates persist. The practical consequence is reduced cold-water solubility, and because interfacial and gelling performance depend first on how much protein dissolves, foaming capacity and heat-set gel strength decline with it. Partial unfolding on its own raises surface hydrophobicity and can improve adsorption at an interface, so solubility is generally the stronger predictor of function than retained tertiary structure, though composition, aggregate state, and interfacial kinetics all contribute.

Alkaline conditions can also oxidize phenolics to quinones that bind protein and discolor the isolate, the dominant discoloration route in phenolic-rich crops such as sunflower, where chlorogenic acid oxidation produces green-colored protein complexes. Above roughly pH 10, and increasingly with temperature and residence time, extraction begins to generate lysinoalanine and racemized residues, making extraction pH a nutritional decision as well as a yield decision. At the 40 °C and 60 min hold used here, lysinoalanine formation is expected to be small even at the upper end of the pH range. Phytate co-precipitates with protein below the pI and carries through into the product. The discarded supernatant often removes the albumin fraction, which in pulses is the sulfur-amino-acid-rich one. For applications that depend on high solubility at neutral pH, clear beverages in particular, IPI is a poor starting point.

*These literature figures were computed almost entirely with a nitrogen-to-protein factor of 6.25; recalculated with the commodity factors this platform applies (see Nitrogen-to-Protein Conversion Factors), the same materials would report roughly 9 to 15% lower.

Equipment
  • Thermomix TM6
  • 1 L centrifuge bottles
  • Centrifuge (SLC-4000 rotor)
  • Stainless steel sieve (200 Mesh)
  • pH meter (Hanna Instruments Foodcare; see Appendix B)
  • Conductivity meter (Thermo Fisher Orion; see Appendix B)
  • Balance (see Appendix B)
Consumables
  • Plant flour, defatted where the commodity requires it, or commercial protein concentrate
  • DI water
  • 1M NaOH
  • 1M HCl
Procedure
  1. Solubilization. Mix 10% (w/w) powder with DI water in the Thermomix for 2000 g total. Disperse at mixing speed 3.0. Adjust to Target pH (determined by Yield Screen) with 1M NaOH. Heat to 40 °C and hold at mixing speed 3.0 for 60 min, readjusting pH every 10 min.
  2. First spin - remove insolubles. Centrifuge mixture for 20 min at 4,690 × g (5000 rpm, SLC-4000 rotor) and 25 °C. Decant supernatant through a sieve. Reserve a 2 mL aliquot of the sieved extract for LECO and SDS-PAGE. Discard the pellet (insolubles).
  3. IPI route - precipitation, second spin. Stir solubilized protein mixture at 900 rpm; slowly add 1M HCl to reach the isoelectric point (literature derived). Stir for 10 min and readjust pH as needed. Centrifuge for 20 min at 4,690 × g (5000 rpm, SLC-4000 rotor) and 25 °C. Discard the supernatant.
  4. Washing. Resuspend pellet in 10x DI water on a wet mass basis of the pellet. Centrifuge for 15 min at 4,690 × g (5000 rpm, SLC-4000 rotor) and 25 °C. Repeat for 2 total washes.
  5. Final extract. Collect the pellet, and record extract mass, pH, and observations. Measure and record the conductivity of the final wash supernatant. Store at 4 °C.
References
  1. McClements, D. J., & Grossmann, L. (2022). Next-Generation Plant-based Foods: Design, Production, and Properties. Springer. https://doi.org/10.1007/978-3-030-96764-2
  2. Cruz-Solis, I., Ibarra-Herrera, C. C., Rocha-Pizaña, M. del R., & Luna-Vital, D. (2023). Alkaline extraction–isoelectric precipitation of plant proteins. In A. J. Hernández-Álvarez, M. Mondor, & M. G. Nosworthy (Eds.), Green Protein Processing Technologies from Plants (pp. 1–29). Springer. https://doi.org/10.1007/978-3-031-16968-7_1
  3. Miranda, C. G., Speranza, P., Kurozawa, L. E., & Sato, A. C. K. (2022). Lentil protein: impact of different extraction methods on structural and functional properties. Heliyon, 8(11), e11775. https://doi.org/10.1016/j.heliyon.2022.e11775
  4. Eckhardt, L., Bu, F., Franczyk, A., & Ismail, B. P. (2024). Hemp (Cannabis sativa L.) protein: Impact of extraction method and cultivar on structure, function, and nutritional quality. Current Research in Food Science, 8, 100746. https://doi.org/10.1016/j.crfs.2024.100746
  5. Boye, J. I., Aksay, S., Roufik, S., et al. (2010). Comparison of the functional properties of pea, chickpea and lentil protein concentrates processed using ultrafiltration and isoelectric precipitation techniques. Food Research International, 43, 537–546. https://doi.org/10.1016/j.foodres.2009.07.021
  6. Yang, J., Zamani, S., Liang, L., & Chen, L. (2021). Extraction methods significantly impact pea protein composition, structure and gelling properties. Food Hydrocolloids, 117, 106678. https://doi.org/10.1016/j.foodhyd.2021.106678
  7. Wildermuth, S. R., Young, E. E., & Were, L. M. (2016). Chlorogenic acid oxidation and its reaction with sunflower proteins to form green-colored complexes. Comprehensive Reviews in Food Science and Food Safety, 15(5), 829–843. https://doi.org/10.1111/1541-4337.12213
  8. Hou, F., Ding, W., Qu, W., et al. (2017). Alkali solution extraction of rice residue protein isolates: influence of alkali concentration on protein functional, structural properties and lysinoalanine formation. Food Chemistry, 218, 207–215. https://doi.org/10.1016/j.foodchem.2016.09.064
  9. Friedman, M. (1999). Chemistry, biochemistry, nutrition, and microbiology of lysinoalanine, lanthionine, and histidinoalanine in food and other proteins. Journal of Agricultural and Food Chemistry, 47(4), 1295–1319. https://doi.org/10.1021/jf981000+
  10. Vogelsang-O'Dwyer, M., Petersen, I. L., Joehnke, M. S., et al. (2020). Comparison of faba bean protein ingredients produced using dry fractionation and isoelectric precipitation: techno-functional, nutritional and environmental performance. Foods, 9, 322. https://doi.org/10.3390/foods9030322
In the PDF, p. 8

Micellar Protein Isolation

Micellar protein isolation exploits the ionic strength dependence of globulin solubility rather than the pH dependence of surface charge, so the protein is never carried through the pH extremes that partially denature an IPI isolate. Extraction is carried out at high ionic strength, typically 0.3 to 1.0 M NaCl at or near the native pH of the dispersion, which for most seed flours is mildly acidic. Across the commodities screened to date, the selected solubilization pH is the unadjusted native pH for the majority, with a minority selecting a mildly alkaline target. At this concentration each protein molecule sits in a dense atmosphere of counterions that screens its charged surface patches and suppresses the long-range electrostatic interactions between molecules. The protein disperses, and insoluble material is removed by centrifugation.

What is reported

  • Solubilization conditions: Initial pH, Target pH, Final pH
  • Solubilized & sieved extract mass (g)
  • Solubilized & sieved extract protein content (%)
  • Extract pH
  • Extract observations: Color, texture
  • Solubilization efficiency (%)
  • Number of washes
  • Final supernatant conductivity (mS/cm)
  • Protein yield

Protein content and protein yield for extracts produced by this route are determined as described under Nitrogen Concentration (Dumas Method). Routine pH, conductivity, and mass measurements follow Appendix B.

Full principle

The solubilized protein is then diluted into cold water. As the salt concentration falls, the counterion atmosphere around each protein thins, and charges that were masked at high ionic strength begin to be impactful at longer range. Protein molecules that had been effectively invisible to one another start to interact, and where the balance of attraction, repulsion, and hydration favors association, the protein self-associates into a protein micellar mass, which settles and is recovered by centrifugation. The name is an analogy rather than a description, since the assembly is amorphous, has no critical micelle concentration, and is held by a balance of electrostatic and hydrophobic contributions rather than by a hydrophobic drive alone. Dilution is also fast relative to conformational and quaternary rearrangement, so the protein meets its neighbors before it can adjust to the new solvent and the product is kinetically trapped. What comes out of the tank depends on the path taken to the final ionic strength and not only on that final value, which is why dilution rate, temperature, and mixing are process variables rather than incidental details. The protocol below fixes the dilution ratio, water temperature, and settling time, but does not currently control pour rate or mixing during dilution; these are a known source of run-to-run variability in the Data Garden's first dataset, but will be controlled moving forward.

Strengths. The protein avoids the pH extremes of IPI, and with them both the covalent damage and the aggregation dynamics those steps impose. Structure is substantially retained, with DSC endotherms close to native in both denaturation temperature and enthalpy where an IPI isolate from the same flour shows a reduced or absent transition. Protein content is high, commonly above 90% on a dry basis in the literature, though as with IPI many figures were computed at 6.25 and would report lower under the commodity factors this platform applies.

Weaknesses. Yield is the principal constraint and is strongly crop-dependent, so the route cannot be assumed to transfer from one seed to another without re-optimization of salt concentration and solubilization pH. The operating optimum generally sits at or near the native pH of the dispersion rather than under alkali, which is consistent with the noncovalent premise of the original method but places the extraction within roughly two units of the globulin isoelectric point and makes the route less mechanistically distinct from isoelectric precipitation than the naming implies. Alkaline solubilization is counterproductive here: it drives more protein into the extract but less of it returns on dilution, since dissociated oligomers micellize poorly, so solubilization yield and recovery yield move in opposite directions. Solubility in low ionic strength water is often poor, and in some preparations lower than that of an IPI isolate from the same flour, which follows from the recovery mechanism, since the process selects for the fraction that is insoluble at low ionic strength and that fraction re-associates on redispersion unless salt is added back. The direction is not universal, and salt-extracted isolates have also been reported as the more soluble of the pair. Micellization also selects compositionally, retaining globulins while albumins are lost to the supernatant, so an MPI isolate is not the same protein population as the flour it came from and any comparison against an IPI isolate is partly compositional rather than purely conformational. The dilution step generates a large volume of dilute salt effluent, which is the main barrier to scale and a heavier burden than the salt itself. Process control is meaningfully harder than IPI, with dilution rate, temperature, hold time, and mixing uniformity all interacting with crop and variety to shift the operating point.

Equipment
  • Thermomix TM6
  • 1 L centrifuge bottles
  • Centrifuge (SLC-4000 rotor)
  • Stainless steel sieve (200 Mesh)
  • pH meter (Hanna Instruments Foodcare; see Appendix B)
  • Conductivity meter (Thermo Fisher Orion; see Appendix B)
  • Balance (see Appendix B)
Consumables
  • Plant flour, defatted where the commodity requires it, or commercial protein concentrate
  • NaCl (Diamond Crystal Kosher Salt, ≥99% NaCl; molarity calculated on that basis)
  • DI water
  • 1M NaOH
Procedure
  1. Solubilization. Mix 10% (w/w) powder with NaCl and DI water in the Thermomix for 2000 g total (salt concentration determined by Yield Screen). Disperse at mixing speed 3.0. Adjust to Target pH (determined by Yield Screen) with 1M NaOH. Heat to 40 °C and hold at mixing speed 3.0 for 60 min, readjusting pH every 10 min.
  2. First spin - remove insolubles. Centrifuge mixture for 20 min at 4,690 × g (5000 rpm, SLC-4000 rotor) and 25 °C. Decant supernatant through a sieve. Reserve a 2 mL aliquot of the sieved extract for LECO and SDS-PAGE. Discard the pellet (insolubles).
  3. MPI route - cold dilution, second spin. Slowly pour the total volume of solubilized protein into 5x that volume of cold DI water (4 °C). Allow protein to settle at 4 °C overnight, typically 18 to 20 hours. Discard the clear top liquid layer. Centrifuge the settled protein for 20 min at 4,690 × g (5000 rpm, SLC-4000 rotor) and 25 °C. Discard the supernatant.
  4. Washing. Resuspend pellet in 10x wet pellet mass (g) of DI water. Centrifuge for 15 min at 4,690 × g (5000 rpm, SLC-4000 rotor) and 25 °C. Repeat washes until final supernatant conductivity is below 0.2 mS/cm (no more than 5 washes).
  5. Final extract. Collect the pellet, and record extract mass, pH, and observations. Store at 4 °C.
References
  1. Murray, E. D., Myers, C. D., Barker, L. D., & Maurice, T. J. (1981). Functional attributes of proteins: a noncovalent approach to processing and utilizing plant proteins. In D. W. Stanley, E. D. Murray, & D. H. Lees (Eds.), Utilization of Protein Resources (pp. 158–176). Westport, CT: Food & Nutrition Press.
  2. Murray, E. D., et al. Protein product and process for preparing same. US Patent 4,169,090 (1979); Soybean protein isolate production. US Patent 4,307,014 (1981).
  3. Ismond, M. A. H., Murray, E. D., & Arntfield, S. D. (1986). The role of non-covalent forces in micelle formation by vicilin from Vicia faba. II. The effect of stabilizing and destabilizing anions on protein interactions. Food Chemistry, 21, 27–46. https://doi.org/10.1016/0308-8146(86)90139-1
  4. Ismond, M. A. H., Murray, E. D., & Arntfield, S. D. (1988). The role of non-covalent forces in micelle formation by vicilin from Vicia faba. III. The effect of urea, guanidine hydrochloride and sucrose on protein interactions. Food Chemistry, 29(3), 189–198. https://doi.org/10.1016/0308-8146(88)90132-X
  5. Tanger, C., Engel, J., & Kulozik, U. (2020). Influence of extraction conditions on the conformational alteration of pea protein extracted from pea flour. Food Hydrocolloids, 107. https://doi.org/10.1016/j.foodhyd.2020.105949
  6. Stone, A. K., Karalash, A., Tyler, R. T., Warkentin, T. D., & Nickerson, M. T. (2015). Functional attributes of pea protein isolates prepared using different extraction methods and cultivars. Food Research International, 76, 31–38. https://doi.org/10.1016/j.foodres.2014.11.017
  7. Yang, J., Zamani, S., Liang, L., & Chen, L. (2021). Extraction methods significantly impact pea protein composition, structure and gelling properties. Food Hydrocolloids, 117, 106678. https://doi.org/10.1016/j.foodhyd.2021.106678
  8. Xu, H.-N., Liu, Y., & Zhang, L. (2015). Salting-out and salting-in: competitive effects of salt on the aggregation behavior of soy protein particles and their emulsifying properties. Soft Matter, 11, 5926–5932. https://doi.org/10.1039/c5sm00954e
  9. Kimura, A., Fukuda, T., Zhang, M., Motoyama, S., Maruyama, N., & Utsumi, S. (2008). Comparison of physicochemical properties of 7S and 11S globulins from pea, fava bean, cowpea, and French bean with those of soybean: French bean 7S globulin exhibits excellent properties. Journal of Agricultural and Food Chemistry, 56(21), 10273–10279. https://doi.org/10.1021/jf801721b
In the PDF, p. 11

Membrane Based Concentration

Membrane-based concentration replaces the pH precipitation used in IPI with ultrafiltration. Protein is solubilized from the flour in water at the target pH selected by the Yield Screen. Insolubles are removed by centrifugation and sieving. The sieved extract is then concentrated across a 5 kDa PES membrane. Solutes above the nominal cutoff are largely retained, including the seed storage globulins, the smaller 2S albumins, soluble polysaccharides, high molecular weight phenolic complexes, and any lipid that survived clarification, while salts, oligosaccharides, and other low molecular weight solutes pass into the permeate. The cutoff is a nominal rejection specification rather than a sharp boundary, so retention near and somewhat above 5 kDa is partial and solute-dependent. This route concentrates anything retained in the soluble fraction of the extract rather than purifying the protein within it. The composition of the final product reflects the composition of the extract it came from.

What is reported

  • Solubilization conditions: Initial pH, Target pH, Final pH
  • Filtration Time
  • Solubilized & sieved extract mass (Initial feed mass) (g)
  • Solubilized & sieved extract protein content (%)
  • Extract pH
  • Extract observations: Color, texture
  • Solubilization efficiency (%)
  • Permeate mass (g)
  • Final Feed Mass (Final extract mass) (g)
  • Concentration factor: Initial feed mass / Final feed mass
  • Final extract pH
  • Protein yield

Protein content and protein yield for extracts produced by this route are determined as described under Nitrogen Concentration (Dumas Method). Routine pH and mass measurements follow Appendix B.

Full principle

Strengths. The protein is never precipitated, so it avoids any acid swing, the neutralization step, and any aggregation dynamics that follow from them. Most of the soluble protein population is retained, including the bulk of the albumin fraction that both precipitation routes tend to discard, which gives a more complete product and a better balanced amino acid profile, since the albumins carry a disproportionate share of the sulfur-containing amino acids. The unit operation runs closed and cold, scales with membrane area at constant flux, and uses a vibration-enhanced module that maintains flux at concentration factors that can foul a conventional crossflow element.

Weaknesses. The route avoids the precipitation excursion but not the solubilization one, since the protein is held at the solubilization pH and elevated temperature for the duration of the extraction, so the structural advantage over IPI is narrower than the absence of a precipitation step implies. Purity is limited by the concentration factor rather than by the membrane, since this route concentrates without diafiltration: permeable solutes are depleted only in proportion to the volume reduction achieved, leaving residual oligosaccharides at a concentration close to that of the feed. The product is a concentrate rather than an isolate, since soluble polysaccharides, phenolic complexes, and residual lipid are retained alongside the protein. Antinutritional factors are carried into the product rather than removed, as lectins and both major trypsin inhibitor classes are retained at this cutoff and phytate is removed only to the extent that it is free rather than protein-bound. Sodium associated with the polyanionic protein is not removed by this route, since there is no wash or diafiltration step, and finishing the run at alkaline pH tends to increase this effect. Capital cost can be high, membranes are a consumable, and flux decline and cleaning remain the governing operating burden at scale.

Equipment
  • Thermomix TM6
  • 1 L centrifuge bottles
  • Centrifuge (SLC-4000 rotor)
  • Stainless steel sieve (200 Mesh)
  • Sani Membranes Vibro-Lab280
  • Snyder Filtration element (MT MAX - PES 5kDa)
  • Peristaltic pump (Watson-Marlow 323U/D)
  • Refrigerator (4-6 °C) for feed reservoir
  • pH meter (Hanna Instruments Foodcare; see Appendix B)
  • Balance (see Appendix B)
Consumables
  • Plant flour, defatted where the commodity requires it, or commercial protein concentrate
  • DI water
  • 1M NaOH
Procedure
  1. Solubilization. Mix 10% (w/w) powder with DI water in the Thermomix for 2000 g total. Disperse at mixing speed 3.0. Adjust to Target pH (determined by Yield Screen) with 1M NaOH. Heat to 40 °C and hold at mixing speed 3.0 for 60 min, readjusting pH every 10 min.
  2. First spin - remove insolubles. Centrifuge mixture for 20 min at 4,690 × g (5000 rpm, SLC-4000 rotor) and 25 °C. Decant supernatant through a sieve. Reserve a 2 mL aliquot of the sieved extract for LECO and SDS-PAGE. Discard the pellet (insolubles).
  3. Ultrafiltration. Record the feed mass. Hold the feed reservoir in a refrigerator set to 4-6 °C, with the feed line passing through the refrigerator wall to the membrane module and the retentate returning to the reservoir, so the bulk of the circulating volume stays chilled for the duration of the run. Concentrate the solubilized protein on the Sani Vibro-Lab280 system to ≥10% protein, with vibration on. Pump at 170 rpm. Set the retentate backpressure valve so that the feed-line gauge reads 1.5-1.75 bar (max 2); with the permeate line fed to its own reservoir. The feed-side pressure is taken as the transmembrane pressure. Sample the retentate periodically and determine protein content by combustion (see Nitrogen Concentration) to establish when the target is reached. Record permeate mass and filtration time at the endpoint. Runs can take up to 24 hours.
  4. Final extract. Collect the retentate and record mass, pH, and observations. Store at 4 °C.
  5. Cleaning. Run the CIP protocol on the membrane element after every run. Store the cleaned element under sterile conditions until next use.
References
  • Boye, J. I., Aksay, S., Roufik, S., Ribéreau, S., Mondor, M., Farnworth, E., & Rajamohamed, S. H. (2010). Comparison of the functional properties of pea, chickpea and lentil protein concentrates processed using ultrafiltration and isoelectric precipitation techniques. Food Research International, 43(2), 537–546. https://doi.org/10.1016/j.foodres.2009.07.021
  • Alonso-Miravalles, L., Jeske, S., Bez, J., et al. (2019). Membrane filtration and isoelectric precipitation technological approaches for the preparation of novel, functional and sustainable protein isolate from lentils. European Food Research and Technology, 245, 1855–1869. https://doi.org/10.1007/s00217-019-03296-y
  • Mondor, M., Aksay, S., Drolet, H., Roufik, S., Farnworth, E., & Boye, J. I. (2009). Influence of processing on composition and antinutritional factors of chickpea protein concentrates produced by isoelectric precipitation and ultrafiltration. Innovative Food Science and Emerging Technologies, 10(3), 342–347. https://doi.org/10.1016/j.ifset.2009.01.007
  • Tanger, C., Engel, J., & Kulozik, U. (2020). Influence of extraction conditions on the conformational alteration of pea protein extracted from pea flour. Food Hydrocolloids, 107.
  • Ali, F., Ippersiel, D., Lamarche, F., & Mondor, M. (2010). Characterization of low-phytate soy protein isolates produced by membrane technologies. Innovative Food Science and Emerging Technologies, 11(1), 162–168. https://doi.org/10.1016/j.ifset.2009.08.004
  • Jaffrin, M. Y. (2008). Dynamic shear-enhanced membrane filtration: a review of rotating disks, rotating membranes and vibrating systems. Journal of Membrane Science, 324(1), 7–25.
  • Akoum, O., Jaffrin, M. Y., & Ding, L. H. (2005). Concentration of total milk proteins by high shear ultrafiltration in a vibrating membrane module. Journal of Membrane Science, 247(1–2), 211–220.
  • Mondor, M., Ippersiel, D., Lamarche, F., & Boye, J. I. (2004). Production of soy protein concentrates using a combination of electroacidification and ultrafiltration. Journal of Agricultural and Food Chemistry, 52(23). https://doi.org/10.1021/jf0400922
  • Souza, P. F. N. (2020). The forgotten 2S albumin proteins: importance, structure, and biotechnological application in agriculture and human health. International Journal of Biological Macromolecules, 164, 4638–4649. https://doi.org/10.1016/j.ijbiomac.2020.09.049
In the PDF, p. 15

Hydration of Commercial Isolates

Hydration

A commercial protein isolate has already been separated from starch and fiber by its manufacturer, so the Data Garden does not extract it. It is characterized as supplied: the powder is dispersed in water to a set protein content and held in the cold until fully hydrated, and the hydrated sample goes through the same assays as an extract. Hydration takes the place of extraction in the workflow, with the same sample handling and coding, and each ingredient is hydrated once. Replicates in its results are technical replicates, as for extracts.

What is reported

  • Protein content of the powder as supplied (combustion nitrogen multiplied by the commodity factor), with its moisture and ash
  • Target protein content of the dispersion and the masses of powder and water used
  • Hydration temperature and time

No mass balance or yield is computed for a hydrated sample, since no protein is separated.

Full principle

A commercial isolate carries its manufacturer's processing history (extraction route, pH treatment, heat, drying), and its functional results reflect that history as well as the crop. Hydrated isolates are therefore compared only with other hydrated commercial ingredients, never with extracts the Data Garden produced from a flour or concentrate.

Procedure
  1. Take the protein content of the powder from its mean combustion nitrogen and the commodity factor (see Nitrogen-to-Protein Conversion Factors).
  2. Weigh the powder and calculate the water needed for 15% protein (w/w): water mass = protein mass × (100 − 15) / 15. Non-protein solids are not counted toward the target.
  3. Disperse the powder in the water and hold at 4 °C for 8 to 10 hours.
  4. Use the hydrated sample directly, diluting it to each assay's loading as for an extract.
In the PDF, p. 19

Composition

The methods in this section determine the protein, moisture, and ash content of the materials in the Data Garden catalog: incoming flours, concentrates and isolates as supplied, the wet extracts produced from flours and concentrates, and hydrated isolates.

Read the rest of this section

All extracts are analyzed in the wet state, without drying, and composition is reported on an as-received wet basis. Drying is itself a processing step that alters solubility and structure, so omitting it keeps the extraction route as the dominant variable distinguishing one extract from another. Where a dry-basis figure is required, whether for comparison across routes or against literature values computed on dried isolates, it is obtained from the wet-basis figure and the moisture content of the same material:

value, dry basis = value, wet basis / (1 − moisture fraction)

Assay loadings are standardized on the wet-basis protein content of the material as collected. Extracts are produced at or above the highest loading any assay requires, so standardization is always a dilution to target. Where a material is run below a method's stated loading, the result is recorded as informational rather than analytical and is not compared against results at target.

Protein content is crude protein calculated from total nitrogen by combustion, using a commodity-specific nitrogen-to-protein conversion factor recorded with every result. The derivation of those factors is described in Nitrogen-to-Protein Conversion Factors.

Composition as reported here is partial. Lipid and carbohydrate are not measured, so the components do not sum and the remainder is not computed by difference, since a figure obtained that way would absorb every unmeasured component together with the accumulated error of the three that are measured. Residual lipid is the more consequential of the two absences. It is retained by the membrane route in particular, it is functionally active at interfaces, and it is present in the oilseed commodities at levels set by a defatting step performed upstream of this platform. Two extracts matched on wet protein content are therefore not matched on composition, and comparisons of functional behavior between routes carry an unmeasured lipid contribution. Lipid determination is a recognized gap and is intended for a later revision of this compendium.

Nitrogen Concentration (Dumas Method)

The Dumas combustion method measures total nitrogen. A weighed sample is combusted at approximately 950 °C in excess oxygen, where organic carbon is oxidized to CO2, hydrogen to H2O, and sample nitrogen is released as a mixture of N2 and nitrogen oxides. The combustion gases are carried in a flowing inert stream through a heated copper reduction stage, which converts NOx back to N2 and removes residual oxygen, and then through scrubbers that strip CO2 and water. What remains is N2 in carrier gas, which a thermal conductivity detector quantifies against a calibration standard, typically EDTA.

Instrument

LECO FP828 Nitrogen Analyzer

What is reported

  • Nitrogen (%)
  • Replicate standard deviation
  • Conversion factor
  • Protein content (%): Nitrogen x conversion factor
Full principle

The detector is not chemically specific. A thermal conductivity cell responds to any gas whose conductivity differs from the carrier, so the specificity of the measurement rests on the reduction and scrubbing stages upstream rather than on the detector itself. The reducing agent is consumed in normal operation and biases results low as it exhausts, often without triggering an instrument error, which is why calibration is verified daily against a certified standard and blanks bracket every run.

The nitrogen result is converted to protein by multiplying by a nitrogen-to-protein conversion factor. The Data Garden uses commodity-specific factors rather than the conventional 6.25, and the factor applied is recorded with every result. See the section on Nitrogen-to-Protein Conversion Factors. The quantity produced is crude protein calculated from total nitrogen. Combustion does not distinguish protein nitrogen from non-protein nitrogen. Free amino acids, nucleotides, nitrate, and other nitrogenous compounds in the sample are counted as protein by this method unless corrected conversion factors are used.

Samples are combusted as received, without drying. The instrument's moisture traps remove combustion water upstream of the detector, so wet extracts, dispersions, and pastes are analyzed directly. Dumas tends to read slightly higher than Kjeldahl on many food matrices, because combustion recovers nitrate and heterocyclic nitrogen that unmodified Kjeldahl digestion does not, which is worth keeping in mind when comparing Data Garden values to older literature.

Equipment
  • LECO FP828 Nitrogen Analyzer
  • Analytical balance
Consumables
  • Tin foil cups (paste/solids)
  • Tin capsules (liquids)
  • EDTA standard (LECO P/N 502-896, 9.55% nitrogen)
  • Compressed oxygen
  • Compressed inert carrier gas (argon)
Procedure
  1. Instrument set up. Check internal parts, filter material, and gas tanks (≥500 psi). Under Diagnostics, run an instrument system check and a leak check.
  2. Startup runs. Load a conditioner sample, 10 blanks, and 3, 150 mg EDTA standards. Confirm the relative standard deviation of the EDTA replicates is ≤2.0%. Individual measurements falling outside ±3% of the mean trigger recalibration before samples are run. Set the Blank and Drift measurements.
  3. Run samples. Run 3 replicates per protein sample. For paste/solids use tin foil cups; for liquids use tin capsules. Target mass varies by sample type (extract, soluble protein, flour), ranging between 150-250 mg.
  4. Shutdown. Run a final blank to confirm no baseline contamination; run a leak check under Diagnostics; turn off gas via software.
References
  • Dee, D. R., & Chang, S. K. C. (2024). Protein Analysis. In B. P. Ismail (Ed.), Nielsen's Food Analysis (6th ed., Ch. 18). Springer. https://doi.org/10.1007/978-3-031-50643-7_18
  • Mauer, L. J. (2024). Moisture and Total Solids Analysis. In B. P. Ismail (Ed.), Nielsen's Food Analysis (6th ed., Ch. 15). Springer. https://doi.org/10.1007/978-3-031-50643-7_15
  • International Organization for Standardization (2008). ISO 16634-1:2008: Food products — Determination of the total nitrogen content by combustion according to the Dumas principle and calculation of the crude protein content — Part 1: Oilseeds and animal feeding stuffs. https://www.iso.org/standard/46328.html.
  • International Organization for Standardization (2016). ISO 16634-2:2016. Food products — Determination of the total nitrogen content by combustion according to the Dumas principle and calculation of the crude protein content — Part 2: cereals, pulses and milled cereal products.
  • AOAC Official Method 992.23. Crude protein in cereal grains and oilseeds, generic combustion method.
  • AOAC Official Method 990.03. Protein (crude) in animal feed, combustion method.
  • Bicsak, R. C. (1993). Comparison of Kjeldahl method for determination of crude protein in cereal grains and oilseeds with generic combustion method: collaborative study. J AOAC Int 76(4), 780-786. https://doi.org/10.1093/jaoac/76.4.780
  • LECO Corporation. (2024). FP828 Nitrogen/Protein Determinator instruction manual. 828 Series Version 3.3x
In the PDF, p. 20

Nitrogen-to-Protein Conversion Factors

The Dumas method of nitrogen measurement results in a raw reading of total detectable nitrogen in a sample. It does not directly measure the concentration of protein in a sample. The protein concentration is extrapolated using our best understanding of the proteins present in a sample. This conversion is made using a nitrogen-to-protein conversion factor (the "N-factor"). The industry standard for labeling purposes is 6.25, meaning that each gram of nitrogen is taken to represent 6.25 g of protein.

Full principle

In reality, a single conversion factor cannot accurately describe the protein concentration of all protein sources, and for plant-based sources of protein, using 6.25 results in a systematic overestimate of the protein in a sample. For example, a legume flour containing 4.0% N yields 25.0% protein at a factor of 6.25, and 22.0% at a more accurate factor of 5.5. Using 6.25 would overstate the protein content by 3 percentage points, or about 14% in relative terms.

Most published protein contents, purities, and functional-assay loadings were computed at 6.25. Values reported by the Data Garden are computed at commodity factors between 5.30 and 5.70 and will therefore sit roughly 9 to 15% below, depending on the commodity factor, a literature figure for the same material. Comparisons across that boundary are not like-for-like unless one side is recalculated.

As our aim is to accurately characterize protein functionality as a result of protein source and protein extraction method, our methods require loading an equal concentration of protein in all assays. The gold standard is to create a sample-specific N-factor using amino acid analysis of the sample itself, together with a measurement of the non-protein nitrogen it carries. Our lab is not equipped with that capability, so the Data Garden uses a hybrid approach, blending AI-assisted literature review with our curation to align on more accurate N-factors.

Conversion factor derivation proceeds in three stages.

Stage one derives a protein-N factor from amino acid composition. We gather literature, focusing especially on those papers which report the amino acid composition of the source material in question. These tables are percent composition tables of amino acids, usually expressed in a value of g amino acid per 100g of sample. Each amino acid contributes nitrogen according to the number of nitrogen atoms it carries and its molecular weight, and contributes mass to the protein chain as a residue, that is, after the loss of one water molecule on peptide bond formation. Literature tables report free amino acid masses recovered after hydrolysis, so the residue correction converts them back to the mass they occupy in the intact protein. The factor is the ratio of total residue mass to total amino acid nitrogen:

k = Σ(residue mass) / Σ(nitrogen mass)

where, for each amino acid at amount A (g/100 g), molecular weight MW, and nitrogen atom count n:

  • nitrogen mass = A × (n × 14.007) / MW
  • residue mass = A × (MW − 18.015) / MW

Constants are the atomic mass of nitrogen (14.007 g/mol) and the molecular weight of water (18.015 g/mol). Molecular weights and nitrogen atom counts are amino acid specific and are listed in Table 2. Nitrogen atom count varies from one for most residues to two for lysine, tryptophan, and cystine, three for histidine, and four for arginine, which is why the factor is composition-dependent rather than universal.

The value this produces is a protein-N factor: the ratio of protein mass to protein-bound nitrogen. It is an upper bound on the applied factor, because Dumas measures total nitrogen rather than protein nitrogen.

Stage two places the factor within a bounded range. The amino-acid-derived factor counts protein-bound nitrogen only, so for a whole flour or seed it is an upper bound: combustion also counts the nitrogen in free amino acids, nucleotides, and other non-protein compounds. Published factors for the same commodity, whether determined on a total-nitrogen basis (Jones, 1941) or compiled as recommendations (Mossé, 1990; Mariotti et al., 2008), sit lower because the material they describe carries that non-protein nitrogen. The adopted factor is placed within that interval, generally at or near the published recommendation for the commodity. Non-protein nitrogen is therefore accounted for by where the factor sits in the interval, rather than by a separately estimated correction.

Stage three is curation. The derived value is cross-checked, where possible, with additional literature sources that report N-conversion factors, as well as other sources of amino acid composition, most often USDA's FoodData Central. At this point, scientists at the Data Garden review the source material to understand and resolve any conflicting values (e.g. hulled vs. dehulled sources, non-protein nitrogen assumptions). We then adopt a conversion factor that more accurately reflects the nitrogen-to-protein ratio of the commodity. The rationale for each N-conversion factor used by the Data Garden is captured, recorded, and updated if any changes are deemed necessary.

Each commodity carries a single factor, set on its flour and applied to the flour, to every extract made from it, and to commercial concentrates and isolates of the same commodity. The factor applied to a given result is recorded with that result. Table 1 gives the commodity-level factor, which is the factor carried on the flour-form ingredient rows for that commodity. Appendix A reproduces the audit trail of factor decisions for those flours.

Limitations. The logical next improvement to N-conversion factors would be to correct for protein purity after isolation. We have decided to leave that for future work as there isn’t enough data in the scientific literature that apply to all three of our extraction techniques to consistently inform our choices. While we are unsatisfied by this outcome, we will revisit the topic when possible.

Conversion factors by commodity

Table 1. Nitrogen-to-protein factor applied as the commodity-level factor in the Data Garden catalog, drawn from the flour-form nitrogen_factor value on the ingredient rows for that commodity (see rationale above). Where two flour ingredients disagree, the range is reported verbatim rather than averaged. The decision history behind each flour factor is recorded in Appendix A.

Commodity N-factor Flour ingredients
Cannabis sativa (Hemp) 5.30 1
Cicer arietinum (Chickpea) 5.40 1
Cucurbita spp. (Pumpkin) 5.50 1
Glycine max (Soybean) 5.50 2
Helianthus annuus (Sunflower) 5.30 1
Lens culinaris (Lentil) 5.40 1
Linum usitatissimum (Flaxseed) 5.70 1
Pisum sativum (Pea) 5.36 1
Sesamum indicum (Sesame) 5.40 1
Vicia faba (Faba Bean) 5.50 2
Lupinus spp. (Lupin) 5.46 0 (isolate only)

Table 2. Nitrogen atom count and molecular weight used for each amino acid in the factor calculation.

Amino acid N atoms MW (g/mol) Amino acid N atoms MW (g/mol)
Alanine 1 89.09 Leucine 1 131.17
Arginine 4 174.20 Lysine 2 146.19
Aspartic acid 1 133.10 Methionine 1 149.21
Cystine 2 240.30 Phenylalanine 1 165.19
Glutamic acid 1 147.13 Proline 1 115.13
Glycine 1 75.07 Serine 1 105.09
Histidine 3 155.16 Threonine 1 119.12
Isoleucine 1 131.17 Tryptophan 2 204.23
Tyrosine 1 181.19
Valine 1 117.15
References
  1. Mariotti, F., Tomé, D., & Mirand, P. P. (2008). Converting Nitrogen into Protein—Beyond 6.25 and Jones' Factors. Critical Reviews in Food Science and Nutrition, 48(2), 177–184. https://doi.org/10.1080/10408390701279749
  2. Sosulski, F. W., & Imafidon, G. I. (1990). Amino Acid Composition and Nitrogen-to-Protein Conversion Factors for Animal and Plant Foods. Journal of Agricultural and Food Chemistry, 38(6), 1351–1356. https://doi.org/10.1021/jf00096a011
  3. Jones, D. B. (1941). Factors for Converting Percentages of Nitrogen in Foods and Feeds into Percentages of Proteins. USDA Circular No. 183 (Slightly Revised). Washington, DC: United States Department of Agriculture.
  4. Mossé, J. (1990). Nitrogen-to-protein conversion factor for ten cereals and six legumes or oilseeds. A reappraisal of its definition and determination. Variation according to species and to seed protein content. Journal of Agricultural and Food Chemistry, 38(1), 18–24. https://doi.org/10.1021/jf00091a004
  5. U.S. Department of Agriculture, Agricultural Research Service. FoodData Central. https://fdc.nal.usda.gov/ (accessed May to September 2026, during the factor decisions recorded in Appendix A; SR Legacy and Foundation Foods releases as cited per ingredient in Appendix A).
In the PDF, p. 22

Moisture and Ash Content

Moisture and ash are gravimetrically determined. Moisture is measured as mass loss on drying at 105 °C overnight for a minimum of 15 hours, allowing the sample to reach a constant weight. The balance records total volatile loss, not water alone, and in protein-rich samples lipid oxidation and Maillard reactions may offset that loss in either direction. Dried isolates are hygroscopic, so crucibles are held hot in the oven until the moment of weighing and are then transferred to the balance and weighed immediately.

Instrument

Muffle Furnace (Across International CF1100)

What is reported

  • Moisture Content (%)
  • Ash Content (Dry Basis) (%)
  • Ash Content (Wet Basis) (%)

Moisture (%) = (wet sample mass − dried sample mass) / wet sample mass × 100
Ash, wet basis (%) = ash mass / wet sample mass × 100
Ash, dry basis (%) = ash mass / dried sample mass × 100

Here the wet basis is the sample as loaded into the crucible, which for extracts is the wet material as collected from the extraction route.

Full principle

Ash is the residue after ignition at 550 °C, taken as an estimate of total mineral content. Carbonates decompose and some minor constituents are lost or altered under these conditions, so the correspondence to the mineral content of the original material is approximate. In isolates the residue includes intrinsic seed minerals, phytate-associated minerals, and salt introduced during solubilization. How much of the last remains depends on how completely it was washed out. Ash therefore reflects both the commodity and the extraction route, and in salt-extracted material the route contribution can dominate. Residual salt is tracked during extraction by the conductivity of the wash supernatant, which is measured at the end of each isolation and, in the micellar route, sets the endpoint for washing. Conductivity reports the ions leaving in the wash water while ash reports what remains in the isolate, so the two are read together: a low final conductivity with a high ash figure points to mineral that is bound rather than free.

Equipment
  • Analytical balance
  • Oven
  • Crucibles
  • Muffle Furnace (Across International CF1100)
Consumables
  • None
Procedure
  1. Prepare crucibles. Crucibles are stored in the drying oven at 105 °C. Transfer to the balance and record initial crucible mass immediately.
  2. Dry samples. Add 2-3 g of sample to the crucible, and record mass. Run in triplicate. Dry at 105 °C for at least 15 hours, or until mass is constant. Transfer directly from the oven to the balance and record final dried mass immediately.
  3. Burn samples. Burn crucibles at 550 °C for 18 hours, or until mass is constant. Return to the 105 °C oven to cool, then transfer to the balance and record final burned mass immediately.
References
  1. Mauer, L. J. (2024). Moisture and Total Solids Analysis. In B. P. Ismail (Ed.), Nielsen's Food Analysis (6th ed., Ch. 15). Springer. https://doi.org/10.1007/978-3-031-50643-7_15
  2. Harris, G. K. (2024). Ash Analysis. In B. P. Ismail (Ed.), Nielsen's Food Analysis (6th ed., Ch. 16). Springer. https://doi.org/10.1007/978-3-031-50643-7_16
In the PDF, p. 26

Mass Balance and Yield Accounting

Nitrogen and mass are recorded at three points in each extraction: the starting powder, the sieved extract after the first spin, and the final extract. The known powder mass and protein content set the protein charged to the batch, and the mass and protein content of each subsequent stream account for what carries forward, so the balance closes by difference without analyzing the discarded streams. Hydrated commercial isolates are not extracted, so no balance or yield is computed for them.

Read the rest of this section

Solubilization efficiency is the protein in the sieved extract as a fraction of the protein in the starting powder. Overall protein yield is the protein in the final extract as a fraction of the protein in the starting powder. Total yield is the protein in the final extract as a fraction of the mass of starting powder charged (overall protein yield × the powder's protein content), the figure a process engineer sizes a batch by. Loss is the remainder. It falls in two places: protein that never entered solution leaves in the first-spin pellet, and protein that dissolved but was not recovered is lost at precipitation or settling, in the washes, or in the membrane permeate. Which of the two dominates varies by route and material.

Each route is solubilized as its own batch and carries its own sieved extract measurement, and all batches for a given material are run on the same day from the same lot. Isoelectric precipitation and membrane concentration run identical solubilization conditions, so their solubilization efficiencies are expected to agree, and the agreement between them is a direct check on batch-to-batch variation. Micellar isolation is solubilized at its own pH and salt concentration and is not comparable at this stage.

Limitations

Loss is a single quantity obtained by difference. Protein may be lost in the insoluble pellet, the precipitation supernatant, during washes, and possibly in the ultrafiltration permeate. This means that mass balance accounting is incomplete. Extraction optimization resides outside of our scope of functionality testing, however this loss reporting does imply actionable process steps for optimization.

In the PDF, p. 28

Structure and Physical State

Protein Fingerprinting (SDS-PAGE)

Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) resolves a protein mixture by subunit molecular weight. SDS coats the protein chain with negative charge proportional to the length of the protein, denaturing it into a uniform-charge-density rod. Under an applied electric field, proteins migrate through a polyacrylamide gel matrix; the smaller they are, the faster they travel through the gel. Separation is by apparent electrophoretic mobility under these denaturing conditions, which tracks molecular weight closely but is not identical to it.

Instrument

Bio-Rad Mini-PROTEAN Tetra Cell

What is reported

  • Annotated gel image
  • Lane intensity profile, expressed in kDa via the ladder calibration
  • Per-band integrated intensity and relative abundance (% of lane)
  • Molecular-weight calibration R²
Full principle

SDS-PAGE reveals the subunit composition of the storage protein fractions. For example, soy 7S β-conglycinin appears as three subunits at approximately 52, 72, and 76 kDa; under reducing conditions, the 11S glycinin runs as acidic (~35 kDa) and basic (~22 kDa) chains. The ratio of bands at the major positions provides a structural fingerprint that can be associated with crop, variety, and isolation route.

Detection uses stain-free chemistry, in which the gel's trihalo compounds are UV-activated and react with tryptophan residues to generate fluorescence. Band intensity therefore scales with the tryptophan content of the protein in that band rather than with its mass, and subunits differing in tryptophan will not give equal signal for equal mass.

Band quantification is performed automatically on ingestion. LIMBBS extracts an intensity profile down each lane, subtracts a linear baseline, and detects bands as peaks in that profile. Ladder bands entered by the operator calibrate a log-linear fit of molecular weight against migration distance, which converts positions to kDa and is reused across every lane on that gel. Each band is reported as an integrated intensity and as a percentage of its lane total. Absolute quantification is not attempted; intensities are compared within a single gel, and because stain-free signal tracks tryptophan rather than mass, relative abundance is a fingerprint metric rather than a mass fraction.

Bands can be assigned to named storage proteins. Each band's calibrated molecular weight is compared with the sequence masses of candidate proteins for the commodity, drawn from UniProt, and scored together with its behavior between the reducing and non-reducing lanes, its expected relative abundance, and published assignments. The best-scoring candidate is recorded with its UniProt accession and a confidence level. A band is named only when the evidence supports the match, and otherwise stays unidentified; a band outside the ladder's resolving range cannot be assigned with high confidence. These are identifications by mass and behavior, not by sequencing.

Equipment
  • Bio-Rad Mini-PROTEAN Tetra Cell
  • Power supply (Bio-Rad PowerPac)
  • Heat block
  • Microcentrifuge
  • Bio-Rad Gel Doc EZ Imager with stain-free tray
  • Image Lab software
Consumables
  • Stain-free precast gels (Bio-Rad 4-20% MP TGX Stain-Free Gel)
  • 10x SDS-PAGE running buffer (250mM Tris-base, 1.923M Glycine, 1% SDS)
  • 4x Laemmli sample buffer (62.5 mM Tris-HCl, 10% glycerol, 1% SDS, 0.005% Bromophenol Blue)
  • 4x Laemmli sample buffer with 400 mM DTT (reducing)
  • BioRad Precision Plus Protein Unstained Standard
Procedure
  1. Prepare samples. Dilute solubilized protein and extract samples to 0.1% protein (w/w) in DI water. For each protein sample, prepare microtube aliquots for reducing and non-reducing conditions: 75 uL 0.1% protein + 25 uL 4x Laemmli buffer (with and without DTT). Heat samples at 95 °C for 5 minutes. Centrifuge for 10 seconds to consolidate any condensed vapor.
  2. Prepare gel. Dilute 10x running buffer to 1x. Load the gel into the cell, and fill the inner and outer chamber with 1x running buffer. Load 15 uL Protein Unstained Standard into Lane 1. Load 5 uL or 10 uL of sample into the lanes (3.75 ug or 7.5 ug protein loaded). Place the lid on the chamber with the red and black marks aligned. Plug the black and red cables into the power supply.
  3. Electrophoresis. On the power supply, set the voltage to 150 V and run time to 1 hour. Allow to run until the blue dye front reaches the bottom of the gel.
  4. Gel imaging. Break the gel out of the casing, and rinse with DI water three times. Transfer gel to the gel imager tray, and follow the Stain-free imaging protocol on Image Lab.
  5. Densitometry. Upload the gel image to LIMBBS with the lane geometry and sample-condition labels. Identify each ladder band on the marker lane and enter its known molecular weight; this calibrates the gel. Review the auto-detected lane boundaries and bands, and adjust if needed, before finalizing the run.
References
  1. Laemmli, U. K. (1970). Cleavage of Structural Proteins during the Assembly of the Head of Bacteriophage T4. Nature, 227(5259), 680–685. https://doi.org/10.1038/227680a0
  2. Dee, D. R., & Chang, S. K. C. (2024). Protein Analysis. In B. P. Ismail (Ed.), Nielsen's Food Analysis (6th ed., Ch. 18). Springer. https://doi.org/10.1007/978-3-031-50643-7_18
  3. Aroma J., Ravneet K., Deciphering food proteins: The applications of SDS-PAGE in food science, Food Bioscience, Volume 66, 2025, https://doi.org/10.1016/j.fbio.2025.106126
  4. Weber, K., & Osborn, M. (1969). The reliability of molecular weight determinations by dodecyl sulfate-polyacrylamide gel electrophoresis. Journal of Biological Chemistry, 244(16), 4406–4412.
  5. Ladner, C. L., Yang, J., Turner, R. J., & Edwards, R. A. (2004). Visible fluorescent detection of proteins in polyacrylamide gels without staining. Analytical Biochemistry, 326(1), 13–20.
In the PDF, p. 29

Thermal Transitions (Differential Scanning Calorimetry)

Differential scanning calorimetry measures the difference in heat flow between a sample and an inert reference as both are heated together on a common thermoelectric platform. When a protein undergoes a thermal transition, most commonly denaturation, heat energy is absorbed by the protein as weak intramolecular bonds are overwhelmed, causing a distinct endothermic event, relative to the reference sample. The result is a thermogram with a peak whose onset, peak, and end temperatures define the transition window, and whose area quantifies the enthalpy of denaturation (ΔH, J/g protein).

Instrument

TA Instruments Discovery DSC 2500

What is reported

  • Temperature onset (°C): Extrapolated onset temperature of the denaturation event.
  • Peak temperature, T_peak / Td (°C): Temperature at maximum heat flow.
  • Temperature endset (°C): Extrapolated endset temperature.
  • ΔH (J/g sample): Enthalpy of unfolding, normalized to the total sample mass loaded.
  • ΔH (J/g protein): Enthalpy of unfolding, normalized to the protein mass in the pan.
  • Sample mass (mg)
  • Sample preparation: Extract vs. 10% protein dispersion
  • Transition detected: 1 when the thermogram carried an integrable endotherm (so Td and ΔH are reported), 0 when the thermogram was read and had none. A run that was not made has no value at all.

ΔH is computed by LIMBBS on ingestion of the raw thermogram, normalized to the total sample mass entered at setup. Where the operator has also integrated the thermogram in the instrument's own software (TA Instruments TRIOS), those values are recorded and shown beside the LIMBBS values as a cross-check; the LIMBBS values are the ones reported. Normalization to protein mass follows:

ΔH (J/g protein) = ΔH (J/g sample) / protein fraction

where the protein fraction is the protein content of the material as loaded, determined by combustion (see Nitrogen Concentration) on the same lot and on the same wet basis. For a dispersion diluted to a target protein concentration, the protein fraction is that target rather than the protein content of the undiluted extract.

Full principle

For seed storage proteins, the thermogram is a useful indicator of native-state preservation. A fully denatured isolate shows little or no endothermic peak; a well-handled isolate can resolve distinct peaks for the storage protein classes present, for example separated 7S and 11S peaks for soy globulins, though peak resolution depends on crop, fraction composition, and how far apart the transitions sit. Comparing pre- and post-process thermograms indicates how much denaturation has occurred during processing. Residual enthalpy is not a direct measure of native fraction: aggregation can suppress the endotherm without complete unfolding, and salt or sugar in the sample can raise both the transition temperature and the apparent enthalpy. Comparisons are meaningful only between samples of comparable composition, protein loading, and hydration.

Samples are run at the pH at which they were collected, which is route-dependent: near the precipitation pH for IPI, near the solubilization pH for MEM, and near the native pH of the dispersion for most MPI extracts. This preserves the material in the state the extraction produced, including the pH history that shaped it, but it means pH is not matched across routes and contributes to the differences between thermograms. The contribution is not symmetric. Thermal stability of seed globulins is highest near the isoelectric point and declines on either side, and alkaline conditions in particular depress the 11S denaturation temperature and can merge the 7S and 11S transitions into a single endotherm. An extract collected at alkaline pH will therefore tend to read as lower in transition temperature and less resolved than the same protein measured at neutral pH, independent of what the extraction did to it. Extract pH is recorded with every thermogram and should be read alongside the result. A pH-matched arm would separate this contribution from the structural one and is a recognized gap. However, the presence or absence of an endotherm is not affected by this, so the use of DSC as an indicator of whether a material retains native structure is unaffected. Transition temperature and enthalpy carry the pH contribution.

Samples are run against an empty reference pan. On this instrument the resulting heat capacity offset is removed in baseline subtraction before peak integration, and water blanks run from 20 to 120 °C in hermetically sealed high-volume pans give a stable, reproducible baseline with no vaporization artifact across the protein denaturation range. Matching a water mass into the reference pan would add a gravimetric matching requirement and a differential pressure risk without improving enthalpy accuracy.

Equipment
  • TA Instruments Discovery DSC 2500
  • RCS 90 cooling system
  • Analytical balance
Consumables
  • High-volume DSC pans (TA P/N 900825.902)
  • High-volume Yellow Die Set (TA P/N 901600.901)
  • Pan lids + black o-rings (per pan)
  • Compressed nitrogen
  • Indium calibration standard
Procedure
  1. DSC setup. Power on the instrument. Open nitrogen gas tank. Verify the regulator reads 15 PSI. Open TRIOS, and connect the instrument. Heat flow selection T4 for high-volume pans. Turn on RCS 90 chiller (20 °C set point). Confirm base purge above 300 mL/min and flow rate 50 mL/min.
  2. Calibration. Verify temperature and enthalpy calibration against an indium standard, confirming the melting onset at 156.6 °C and the heat of fusion at 28.71 J/g. Recalibrate if either falls outside tolerance (this is not performed every run).
  3. Sample preparation. For protein extract, test sample as is and diluted to 10% protein in DI water. For powders, dilute to 10% protein in DI water. Stir dispersions for 1 hour at room temperature. Assemble the high-volume pans with 30 mg of sample, seal with the yellow die set, and record the exact mass loaded. Sample masses outside 2 to 30 mg are flagged and no enthalpy is computed.
  4. Create sample runs. Set up sample runs with the following parameters: Pan Type: High Volume Pan. Test: Ramp. Heating Rate: 5 °C/min. Final Temperature: 120 °C. Start Temperature: 20 °C. Enter the recorded sample mass for each pan.
  5. Run samples. Place an empty reference pan in Position 1. Load sample pans into corresponding grid positions. Click Start. The auto-sampler places pans into the cell and runs the ramp. Run each sample in triplicate.
  6. Data quality check. Confirm a stable baseline on either side of the transition and confirm the peak returns to baseline before 120 °C. Peaks are detected in a 50 to 110 °C window, and onset and endset are clipped to that window; a transition that has not closed by 110 °C is reported as truncated. Compare replicate onset and enthalpy values before accepting the run.
  7. Shut down. Wait for cool-down; RCS auto-off. Flip DSC power switch off. Close nitrogen gas tank.
References
  1. Wang, X. S., Tang, C. H., Yang, X. Q., & Gao, W. R. (2008). Characterization, amino acid composition and in vitro digestibility of hemp (Cannabis sativa L.) proteins. Food Chemistry, 107(1), 11–18. https://doi.org/10.1016/j.foodchem.2007.06.064
  2. Schmidt, S. J., & Thomas, L. C. (2024). Thermal Analysis. In B. P. Ismail (Ed.), Nielsen's Food Analysis (6th ed., Ch. 30). Springer. https://doi.org/10.1007/978-3-031-50643-7_30
  3. Fitzsimons, S. M., Mulvihill, D. M., & Morris, E. R. (2007). Denaturation and aggregation processes in thermal gelation of whey proteins resolved by differential scanning calorimetry. Food Hydrocolloids, 21(4), 638–644. https://doi.org/10.1016/j.foodhyd.2006.07.007
In the PDF, p. 31

Particle Size (DLS)

Dynamic light scattering measures particle size from the time-resolved fluctuations of laser light scattering by particles in Brownian motion. Faster fluctuations correspond to smaller, more rapidly diffusing particles. An autocorrelation function is applied to yield a translational diffusion coefficient, which the Stokes-Einstein relation converts to a hydrodynamic diameter, that is, the diameter of an idealized sphere diffusing at the same rate. Proteins are not spheres, so the reported size is an equivalent sphere rather than a true physical dimension. The relationship between the diffusion coefficient and the hydrodynamic diameter depends on the temperature and viscosity of the dispersant.

Instrument

Anton Paar Litesizer 700

What is reported

  • Solvent: Phosphate buffer (pH 7.0) or DI water
  • Protein concentration (%)
  • Z-average diameter (nm)
  • Polydispersity index (PDI)
  • Size distribution plots: Intensity-weighted, number-weighted, and volume-weighted
  • Measured pH of the DI water arm
Full principle

Scattering intensity scales with roughly the sixth power of particle diameter for particles small relative to the incident wavelength. This means that larger particles dominate the signal. A small population of aggregates or a single dust particle can obscure smaller, more plentiful particles. Z-average and polydispersity index (PDI) are cumulants outputs and are meaningful only for narrow, monomodal distributions; intensity distributions come from a separate distribution analysis, and volume- and number-weighted distributions are model-derived transformations of that intensity result rather than independent measurements. DLS cannot resolve populations differing by less than about threefold in diameter.

Because the intensity weighting is so strongly biased toward large particles, this method does not attempt to characterize aggregates. Samples are filtered at 0.45 µm before measurement, and the reported distribution describes the sub-micron protein population that passes the filter. The intent is to compare how extraction routes shape that population across otherwise identical materials, not to quantify the aggregate load, which would overwhelm any signal from the smaller species. Two consequences follow. The cutoff is nominal rather than sharp, so some material above 450 nm passes and the upper bound of the measured population is not precisely defined. And nylon membranes adsorb protein, so the filtrate is depleted relative to the sample by an amount that is not measured. Re-aggregation post filtration is possible, but infrequent due to low protein concentration.

For protein isolates dispersed in water, DLS as run here reports the size distribution of the filterable protein population. A well-handled globulin isolate typically gives a narrow monomodal peak in the low tens of nanometers; denaturation and aggregation broaden the distribution and introduce larger populations within the filterable range. Apparent size depends strongly on pH and ionic strength, so the buffered arm fixes both across samples.

Equipment
  • Anton Paar Litesizer 700
  • pH meter (Hanna Instruments Foodcare; see Appendix B)
Consumables
  • Disposable cuvettes
  • Syringe
  • 0.45 μm filters (Biomed Solutions Nylon)
  • 0.1 M phosphate buffer (pH 7.0)
  • DI water
Procedure
  1. Dilute samples. Standardize samples to 1% protein (w/w) in DI water and 0.1 M phosphate buffer (pH 7.0). Adjust the phosphate buffer sample to pH 7.0 ± 0.1 with 1 M NaOH or 1 M HCl. Leave the DI water sample unadjusted, and record its measured pH. Stir dispersions for 30 min at room temperature, and readjust pH as needed.
  2. Further dilutions - Dilute each sample to 0.1% and 0.05% protein with their respective solvent.
  3. DLS set up. Switch on the DLS at least 10 min before measurements. Set the following input parameters - Measurement cell: Disposable. Measurement Angle: Automatic. Target Temperature: 22 °C. Analysis Mode: General. Cumulant Model: Advanced. Quality: Mode Automatic. Laser Attenuation: Mode Automatic. Focus: Mode Automatic. Material Name: Proteins (refractive index=1.45). Solvent: Water. Water optical and viscosity properties are used for both arms; at 0.1 M phosphate the resulting viscosity error is small relative to the differences this method resolves.
  4. Prepare cuvette. Wearing powder-free gloves, use a syringe and a 0.45 µm filter to filter 1 mL of solution into a disposable cuvette. Discard the first 0.5 mL through a fresh filter before collecting, to reduce depletion by adsorption to the membrane.
  5. Measure. Allow dispersions to reach room temperature before measurement. Insert the cuvette into the instrument chamber, and start measurement. Test 0.1% and 0.05% dilutions for phosphate buffer and DI water samples. Run three replicates each. Confirm that the recorded actual temperature matches the 22 °C setpoint during measurement.
  6. Data quality check. Intensity trace stable, no large peaks or drift. Mean scattering intensity in 20–700 kcounts/s. If intensity falls below range at 0.05%, report the 0.1% result.
References
  1. Amat, T., Assifaoui, A., Buczkowski, J., Silva, J., Schmitt, C., Saurel, R. (2025) Formation of insoluble complexes upon calcium addition to total pea globulins or isolated 7S and 11S fractions: Impact of pH and intrinsic phytic acid content, Food Hydrocolloids, Volume 168, https://doi.org/10.1016/j.foodhyd.2025.111560
  2. Stetefeld, J., McKenna, S. A., & Patel, T. R. (2016). Dynamic light scattering: a practical guide and applications in biomedical sciences. Biophysical Reviews, 8(4), 409–427. https://doi.org/10.1007/s12551-016-0218-6
  3. International Organization for Standardization (2017). ISO 22412:2017. Particle size analysis — Dynamic light scattering (DLS). https://www.iso.org/standard/65410.html
In the PDF, p. 34

Functional Properties

Solubility

Solubility is the starting point for every other functional property. A protein must enter into solution to adsorb at interfaces, emulsify, foam, or form a fine-stranded gel network, so solubility both predicts and constrains most of what follows. Insoluble protein particles can still hydrate, swell, and immobilize water, so the constraint is strong but not absolute.

Instrument

BMG Labtech FLUOstar Omega

What is reported

  • Solubility (%): Supernatant Protein Content (mg/mL) / Initial Protein Content (mg/mL) x 100
  • Protein before and after the spin (mg/mL): each Bradford reading multiplied by the dilution recorded for that well, so the two are on the same footing; the raw in-well readings are not comparable, since the pre-spin sample is read at 25 to 100x and the supernatant at 1 to 25x
  • pH
  • Conductivity (mS/cm)

Solubility can read above 100%. The two protein contents come from different dilutions and different standard curves (the pre-spin sample in 0.1 M NaOH, the supernatant in water), and the supernatant holds a different mix of proteins than the whole dispersion, which can bind the dye more strongly; the ratio carries the combined error of both readings, and Bradford's response differs between crops. Values are published as measured in the dataset downloads. In the Data Garden portal, charts, tables and the API show any solubility above 100% as 100%; the clip is a display convention, not a correction, and it is stated on each chart. Of the extracts in release 2026.1, six read above 100%, all within 10% of it.

Full principle

The material is dispersed at a defined concentration in water, adjusted to a set pH, stirred for a fixed hydration time, and centrifuged to sediment insoluble material. Protein is quantified by Bradford assay in two places: an aliquot of the whole dispersion diluted in alkali, which is taken to solubilize the full protein content, and the supernatant after centrifugation. Solubility is the ratio of the two. Both are read against BSA standard curves prepared in the matching matrix, alkaline for the whole dispersion and aqueous for the supernatant, which corrects for the pH-dependent color response of the Coomassie dye. It does not correct for the dye's unequal response to different proteins: BSA is the calibrant in both curves, and centrifugation fractionates the sample, so the numerator and denominator can carry different protein populations. The residual bias this introduces is largest near the isoelectric point, where the separation between soluble and pelleting protein is most selective.

What the method measures is the fraction of protein that does not pellet under the stated centrifugal conditions, which is an operational quantity rather than thermodynamic solubility. The measurement separates what stays suspended from what pellets, so centrifugal force, spin duration, dispersion concentration, hydration time, and temperature all shift the result. Values are comparable only between samples run under identical conditions.

Running the test across a pH series gives a solubility-pH profile. Plant globulins show a solubility minimum near their isoelectric point, typically pH 4 to 5.5, where net charge and electrostatic repulsion are lowest, and rise toward maxima at neutral and alkaline pH. Ionic strength moves the profile in both directions, since modest salt increases solubility near the isoelectric point while high salt reduces it, so residual process salt is read alongside these results. Note that this assay runs in DI water with no added electrolyte, while the interfacial and hydrophobicity assays run in 0.1 M phosphate. Sections that direct the reader to interpret their results alongside solubility are therefore comparing across a difference in ionic strength, which should be kept in view.

Both protein contents are Bradford values calibrated against BSA. The 1% loading that produced the dispersion is a combustion-derived value (see Nitrogen Concentration), so the ratio is internally consistent even though the two quantifications are not interchangeable. Replicates are triplicate wells from a single dispersion per pH, so the reported dispersion is analytical rather than preparative.

Equipment
  • BMG Labtech FLUOstar Omega Microplate Reader
  • Centrifuge (Thermo Scientific TX-400 Rotor)
  • pH meter (Hanna Instruments Foodcare; see Appendix B)
  • Conductivity meter (Thermo Fisher Orion; see Appendix B)
Consumables
  • DI water
  • 15 mL centrifuge tubes
  • 0.1 M NaOH/ 0.1M HCl
  • 96-well plate
  • Bradford reagent
  • BSA protein standard (2mg/mL)
Procedure
  1. Dilute samples. Standardize samples to 1% protein (w/w) with DI water. Portion 10 mL of dispersion into 5, 15mL tubes. Adjust pH to 4, 5, 6, 7, and 8 with 0.1M NaOH/HCl. Stir dispersions for 30 min at room temperature, and readjust pH as needed. Measure conductivity (mS/cm) of dispersions.
  2. Centrifuge. Collect 1 mL aliquot of each sample before centrifuging to determine initial protein content via Bradford. Draw this aliquot while the dispersion is stirring so that suspended material is sampled representatively. Centrifuge samples 25 min at 1,690 × g (3000 rpm, TX-400 rotor) and 25 °C. Immediately after spinning, collect 1mL of supernatant from each sample.
  3. Create standards. Prepare standard curve concentrations of BSA in DI water and 0.1M NaOH (25, 125, 250, 500, 750, 1000 ug/mL). Blanks are prepared in the same two matrices.
  4. Dilution preparation. For pre-spin 1% samples, dilute in 0.1M NaOH within the range of the standard curve, typically 50x. For post-spin supernatants, dilute in DI water within the working range, typically 1x - 25x.
  5. Plate preparation. Pipette 10 uL of the pre- and post-spin samples in triplicate into a 96-well plate. Pipette 10 uL of DI water and 0.1 M NaOH for blank measurements. Pipette 10 uL for each concentration of the DI water and NaOH BSA standards curves.
  6. Measurement. Add 200 uL Bradford reagent to each sample well. Ensure reagent is at room temperature since the reaction is temperature dependent. Allow the plate to incubate for 10 minutes before measuring absorbance at 595 nm.
  7. Data quality check. Check that the standard curves follow a second order polynomial fit. Ensure the protein sample measurements fall within their respective standard curve (0.1M NaOH for pre-spin and DI water for post-spin samples). Multiply the measured protein concentration by the dilution factor to determine the protein content of each solution.
References
  1. Bradford, M. M. (1976). A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Analytical Biochemistry, 72(1–2), 248–254. https://doi.org/10.1016/0003-2697(76)90527-3.
  2. Caldeira, R. F., Lucas, Tatiana, Wilhelm, A. E., Daniela, Galdeano, M. C., Felberg, I., Rosemar Antoniassi, Mellinger, C. G., & Lima, J. R. (2025). Chickpea processing to obtain a protein ingredient for the plant-based market. Frontiers in Food Science and Technology, 5. https://doi.org/10.3389/frfst.2025.1547684
  3. Ozkan, G., Tataroglu, P., Gulec, S., & Capanoglu, E. (2024). Modification of pea protein isolates by high-intensity ultrasonication: Functional, structural and nutritional properties. Food Chemistry Advances, 5, 100793. https://doi.org/10.1016/j.focha.2024.100793
  4. Dent, T., LeMinh, A., & Farnaz Maleky. (2024). Comparison of Colorimetric Methods for Measuring the Solubility of Legume Proteins. Gels, 10(9), 551–551. https://doi.org/10.3390/gels10090551
In the PDF, p. 37

Surface Hydrophobicity

Surface hydrophobicity describes how much nonpolar surface a protein presents to the surrounding water. It is a structural property with functional consequences: exposed hydrophobic surface contributes to adsorption at air-water and oil-water interfaces, and to the hydrophobic association that builds gel networks. It is one contributor among several rather than a complete account of emulsifying, foaming, or gelling behavior.

Instrument

BMG Labtech FLUOstar Omega

What is reported

  • Net RFI vs. Protein Concentration (mg/mL) plot
  • Slope: Measurement of surface hydrophobicity
  • Instrument gain
  • R² of the fitted linear region
Full principle

The assay uses ANS (8-anilino-1-naphthalenesulfonate), an extrinsic fluorescent probe that is weakly fluorescent in water but fluoresces strongly when bound to nonpolar sites on a protein surface. Fluorescence is measured across a protein dilution series at fixed probe concentration, and the surface hydrophobicity index S₀ is taken as the initial slope of net fluorescence against protein concentration. Net fluorescence subtracts both a probe-only blank, which removes background fluorescence from unbound ANS, and a protein-only blank at each concentration, which removes elastic light scatter and any sample autofluorescence reaching the emission detector. Measurements are made on the FLUOstar Omega with excitation at 355 nm and emission at 460 nm.

The protein-only blank is the primary control for turbidity. It contains identical sample composition and identical particulate load minus the probe, so scatter reaching the detector is subtracted at each dilution point. Residual inner-filter attenuation is managed by fitting only the linear region, since the highest-concentration points where turbidity dominates are excluded from the fit. Absorbance at 390 or 600 nm is not currently captured as an ancillary turbidity flag.

Only the initial linear portion of the curve is used. At higher protein concentrations the relationship flattens, principally through inner filter attenuation of both excitation and emission and through increasing turbidity, so the dilution ladder is chosen to sit within the linear region and linearity is confirmed before the slope is taken.

ANS is described as an aromatic probe, following the convention that distinguishes it from aliphatic probes such as cis-parinaric acid, and the two do not always rank proteins the same way. ANS also carries a negative charge, so binding has an electrostatic component in addition to a hydrophobic one and is influenced by accessible cationic residues. S₀ therefore depends on pH and ionic strength as well as on hydrophobic surface, and dispersion conditions are held constant across samples.

S₀ is method-dependent and expressed in arbitrary units. Probe concentration, buffer, pH, temperature, and instrument gain all affect the value, so comparisons are valid within this protocol and not against published absolute values. The dilution ladder is prepared from an unclarified dispersion and the slope is taken against nominal total protein, so S₀ as constructed is a joint function of surface hydrophobicity and of how much of that protein is in solution. Results are interpreted together with solubility measured on the same material.

Interpretation is not monotonic with processing severity. Mild denaturation exposes previously buried hydrophobic surface and raises S₀, while more extensive denaturation drives aggregation that buries that surface again and lowers it. A low S₀ can therefore mean either well-preserved native structure or aggregated material, which is why this measurement is read alongside DSC and particle size rather than on its own.

Equipment
  • BMG Labtech FLUOstar Omega
Consumables
  • 0.1 M phosphate buffer (pH 7.0)
  • 1 M NaOH/ 1M HCl
  • Aluminum foil
  • 2 mL microtubes
  • Black 96-well plate
  • 8-anilino-1-naphthalenesulfonate (8-ANS)
Procedure
  1. Dilute samples. Standardize samples to 1% protein (w/w) at pH 7.0 ± 0.1 with 0.1 M phosphate buffer (pH 7.0). Adjust with 1 M NaOH or 1 M HCl. Stir dispersions for 30 min at room temperature, and readjust pH as needed.
  2. Prepare ANS. Prepare 2.0mM ANS in 0.1M phosphate buffer (pH 7). Allow to mix for 1 hour at room temperature before using. Wrap in aluminum foil to protect the light-sensitive solution.
  3. Dilution Ladder. Using the 1% protein sample, dilute to 0.1% protein (1.0 mg/mL) in 0.1 M phosphate buffer (pH 7). From that stock, prepare a two-fold serial dilution in the microtubes: 0.5, 0.25, 0.125, 0.0625, 0.031, and 0.016 mg/mL.
  4. Plate Setup. Using a black 96-well plate, pipette 40uL of ANS probe in duplicate for each dilution. Pipette 200 µL of each protein dilution into the probe wells. Prepare two sets of blanks in duplicate: protein blanks, containing 200 µL of each protein dilution plus 40 µL of 0.1 M phosphate buffer in place of the probe; and probe blanks, containing 200 µL of 0.1 M phosphate buffer plus 40 µL of ANS. Matching the buffer volume keeps well volume and path length identical across probe wells and blanks, so that the in-well protein concentration is the same in both.
  5. Measurement. Place the plate into the FLUOstar Omega, and allow samples to incubate in darkness for 15 minutes. Measure fluorescent intensity with 355 nm excitation and 460 nm emission. Use the same gain across every plate in a comparison set, and record the gain with the result.
  6. Data Quality Check. Net RFI = (Sample_probe) − (Protein_blank) − (Probe_blank). Plot Net RFI against the in-well protein concentration, which is the ladder value multiplied by 200/240. Fit the initial linear region and confirm linearity before taking the slope.
References
  • Kato, A., & Nakai, S. (1980). Hydrophobicity determined by a fluorescence probe method and its correlation with surface properties of proteins. Biochimica et Biophysica Acta, 624(1), 13–20. https://doi.org/10.1016/0005-2795(80)90220-2
  • Townsend, A.-A., & Nakai, S. (1983). Relationships between hydrophobicity and foaming characteristics of food proteins. Journal of Food Science, 48, 588–594. https://doi.org/10.1111/j.1365-2621.1983.tb10796.x
  • Nakai, S. (2003). Measurement of protein hydrophobicity. Current Protocols in Food Analytical Chemistry, 9(1), B5.2.1–B5.2.13. https://doi.org/10.1002/0471142913.fab0502s09
  • Idowu, A. O., Alashi, A. M., Nwachukwu, I. D., Fagbemi, T. N., & Aluko, R. E. (2021). Functional properties of sesame (Sesamum indicum Linn) seed protein fractions. Food Production, Processing and Nutrition, 3, Article 4. https://doi.org/10.1186/s43014-020-00047-5
  • Kralik, B., Krentz, A., Creek, M., Saeidy, S., & Ismail, B. P. (2025). Structure-function guided production of oat and chickpea protein isolates in comparison to pea protein isolate following mild extraction conditions. Applied Food Research, 5(2), Article 101440. https://doi.org/10.1016/j.afres.2025.101440
In the PDF, p. 39

Emulsion Capacity

An emulsion is a dispersion of oil droplets in an aqueous phase. It is thermodynamically unstable, so its working lifetime depends on the interfacial film of adsorbed protein that coats the droplets. Protein lowers interfacial tension during droplet formation, a thin viscoelastic film forms around the droplet that resists coalescence and is reinforced by electrostatic and steric repulsion between coated droplets.

Instrument

None. This method uses a homogenizer, water bath, centrifuge, and ruler, all listed under Equipment.

What is reported

  • Emulsion Capacity (%): Emulsion height (mm) / Total height (mm) x 100
  • Emulsion Stability (%): Heated emulsion height (mm) / Emulsion height (mm) x 100

Emulsion Stability here is the heated-layer ratio, following the convention used in the plant protein literature. It is distinct from the Turbiscan Stability Index reported under Emulsion Kinetics, which describes a different emulsion at a different oil load and runs in the opposite direction.

Full principle

Emulsion capacity is the emulsified layer as a fraction of the total column immediately after formation and centrifugation, measured at a high oil load (60% v/v). It is a phase-height index at a single fixed oil load rather than a titration to phase inversion, so it ranks samples by how much of the tube they converted to emulsion under identical conditions and is not comparable to literature emulsion capacities expressed as milliliters of oil per gram protein. Layer height also cannot separate packed droplet fraction from trapped aqueous phase, flocculated but uncoalesced droplets, or unemulsified oil. Emulsion stability is the same measurement repeated after heating, expressed as the fraction of the initial emulsion layer that survives, reflecting how well the interfacial film survives thermal stress, which unfolds adsorbed protein further and can either strengthen the film through interfacial cross-linking or promote coalescence.

Unlike foaming, emulsification does not require extensive unfolding on adsorption. The oil-water interface is less hydrophobic than the air-water interface, so compact globular proteins adsorb and perform well without substantial rearrangement, and emulsion capacity often tracks exposed hydrophobic surface. Results are read alongside surface hydrophobicity and solubility from the same material, since dissolved protein adsorbs most readily, though colloidal protein particles present in these unclarified dispersions can also stabilize an interface.

Emulsion measurements are method-dependent. Homogenization energy and duration, oil fraction, oil type, protein concentration, pH, ionic strength, and centrifugation conditions all shift the result, so values are comparable only between samples run under identical conditions. The result is an accelerated endpoint under centrifugal field rather than a spontaneous separation. This method reports endpoint behavior at high oil load; the rate and mechanism of early destabilization at low oil load are measured separately, in Emulsion Kinetics. Emulsions are formed with a rotor-stator homogenizer, which suits the small volumes of this screen, rather than a high-pressure homogenizer. Droplets formed this way are comparatively large, typically above one micron, so the emulsions cream even when the droplets resist flocculation and coalescence; droplet size is not measured. Because every material is emulsified under the same conditions, the method resolves differences between materials, while absolute emulsion stability depends on the homogenization process. Beverage applications, where creaming governs shelf life, call for high-pressure homogenization, which is planned for the milk application work.

Equipment
  • Rotor-stator homogenizer (PRO Scientific PRO25D, 10 mm probe)
  • Water bath
  • Centrifuge (Thermo Scientific TX-400 Rotor)
  • Ruler
Consumables
  • DI water
  • 0.1 M phosphate buffer (pH 7.0)
  • 1 M NaOH / 1 M HCl
  • 15 mL centrifuge tubes
  • Soybean oil (Stratas Foods, 100% soybean oil)
Procedure
  1. Dilute samples. Standardize samples to 1% protein (w/w) at pH 7.0 ± 0.1 with 0.1 M phosphate buffer (pH 7.0). Adjust with 1 M NaOH or 1 M HCl. Stir dispersions for 30 min at room temperature, and readjust pH as needed.
  2. Prepare emulsions. In a 15 mL centrifuge tube, add 3.2 mL 1% protein dispersion + 4.8 mL soybean oil (60% v/v oil). Before homogenization, shake the tube to break up large oil droplets at the top. Homogenize for 2 min at 10,500 rpm, oscillating the tube from the bottom to the 5mL mark. Avoid crossing the air/surface interface to reduce aeration. Create 3 replicates per sample.
  3. Centrifuge. Centrifuge samples for 5 min at 1,690 × g (3000 rpm, TX-400 rotor) and 20 °C. Measure the height of the emulsified layer. Take an average, as the height may vary around the circumference. Record the Total Height (mm) of the mixture.
  4. Heat. Heat samples upright in an 80 °C water bath for 30 min, and allow to cool to room temperature. Centrifuge for 5 min at 1,690 × g (3000 rpm, TX-400 rotor) and 20 °C. Measure the height of the Heated Emulsion layer (mm).
References
  • Yasumatsu, K., Sawada, K., Moritaka, S., Misaki, M., Toda, J., Wada, T., & Ishii, K. (1972). Whipping and Emulsifying Properties of Soybean Products. Agricultural and Biological Chemistry, 36(5), 719–727. https://doi.org/10.1080/00021369.1972.10860321
  • Ozkan, G., Tataroglu, P., Gulec, S., & Capanoglu, E. (2024). Modification of pea protein isolates by high-intensity ultrasonication: Functional, structural and nutritional properties. Food Chemistry Advances, 5, 100793. https://doi.org/10.1016/j.focha.2024.100793
In the PDF, p. 42

Emulsion Kinetics

Emulsions are thermodynamically unstable, so the useful question is not whether they separate but how quickly and by which route.

Instrument

Turbiscan Tower (Formulaction)

What is reported

  • Turbiscan Stability Index (TSI): sum of scan-to-scan changes in backscattering across the sample height, accumulated over the run and reported at the 1 h endpoint
  • Creaming index (%): the change in backscattering between the last scan and the first, averaged over the top 20% of the sample height, minus the same change over the bottom 20%. A positive value means material moved toward the top of the vial (creaming), a negative one toward the bottom. It is a directional drift indicator computed by LIMBBS from the profiles, not the classical creaming index (serum layer height as a fraction of total height), which this geometry does not resolve.
  • Backscattering profile
Full principle

The emulsion is prepared directly in a scanning vial and profiled undiluted along its full height by near-infrared backscattering and transmission, repeatedly over one hour. Because each scan is spatially resolved, the location of change in the profile carries information about mechanism. Migration, meaning creaming or sedimentation, appears as change confined to the top and bottom zones with a stable middle. Change in droplet size, by coalescence or flocculation, appears as drift in backscattering through the middle zone at fixed position. The distinction is drawn without disturbing the sample, but it is indicative rather than definitive: backscattering responds to both droplet size and local concentration, coalescence and flocculation are not separable from one another by this signal, and the two mechanisms can proceed at once.

Results are reported as the Turbiscan Stability Index, a cumulative sum of scan-to-scan differences across the full sample height, together with the backscattering profile. Higher TSI means greater change and therefore lower stability, the opposite direction to the other functional metrics in this compendium. The one-hour window captures early destabilization kinetics rather than shelf stability, and the profile reports relative change in droplet population rather than absolute droplet size, which is not measured. Backscattering also depends on the droplet size distribution the emulsion started with, which differs between samples, so TSI compares the extent of change in populations that were not identical to begin with.

Stability depends principally on the mechanical properties of the interfacial film and on the repulsion between coated droplets, rather than on how efficiently the protein adsorbed in the first place, though adsorption kinetics and interfacial coverage feed into how complete that film is. A protein that emulsifies readily may still give an unstable emulsion if its film is thin or if charge screening lets droplets approach closely, which is why formation and persistence are measured separately.

Values are method-dependent. Oil fraction, homogenization energy, temperature, protein concentration, pH, and ionic strength all shift them, so comparisons are valid only between samples run under identical conditions. This method uses a lower oil load (20% v/v) and different vessel geometry than Emulsion Capacity, so the two describe different emulsions rather than two views of one.

Equipment
  • Turbiscan Tower (Formulaction)
  • Turbiscan vials
  • Rotor-stator homogenizer (PRO Scientific PRO25D, 10 mm probe)
  • Homogenizer stand
Consumables
  • 0.1 M phosphate buffer (pH 7.0)
  • 1 M NaOH / 1 M HCl
  • Soybean oil (Stratas Foods, 100% soybean oil)
Procedure
  1. Dilute samples. Standardize samples to 1% protein (w/w) at pH 7.0 ± 0.1 with 0.1 M phosphate buffer (pH 7.0). Adjust with 1 M NaOH or 1 M HCl. Stir dispersions for 30 min at room temperature, and readjust pH as needed.
  2. Turbiscan setup. Set Temperature Control to 22 °C, and allow the instrument to equilibrate before testing. In the Towersoft software, set up the sample name and analysis sequence for Emulsion Kinetics: Starts immediately, 13 scans every 5 min (1 h). As for foaming, the realized interval can exceed the set one, and elapsed time is taken from the instrument's time stamps.
  3. Emulsify solution. In a Turbiscan vial, add 9.6 mL 1% protein dispersion + 2.4 mL soybean oil (20% v/v oil). Use a stand to position the vial so the tip of the homogenizer probe sits a few millimeters above the bottom of the vial, centered in the vial cross-section. Homogenize the sample for 2 min at 10,000 rpm, keeping the vial stationary.
  4. Load sample. Immediately cap the vial, and place it into the Turbiscan tower for measurement. The measurement will begin when the door closes. Load within 60 s of the end of homogenization, since the first scan sets the reference against which all subsequent change is measured. Repeat in triplicate.
  5. Sample removal. When samples are done running, ensure the green bars are shown before opening the doors and removing the vials. When finished, turn off the Temperature Control.
  6. Data quality check. Check that the total sample height is around 25 mm. Check the auto-detected meniscus, and adjust height as needed. The bottom of the vial should be around 5mm and the meniscus should be at the top inflection point of backscattering from the first timepoint.
References
In the PDF, p. 44

Foaming Capacity

A protein-stabilized foam is a dispersion of gas bubbles in an aqueous protein solution, where the bubble surfaces are coated by an interfacial film of adsorbed and partially unfolded protein. Two distinct properties govern performance: foaming capacity, the volume of gas incorporated under defined whipping or sparging conditions, and foam stability, the resistance of the foam to collapse by drainage, Ostwald ripening, and coalescence.

Instrument

Turbiscan Tower (Formulaction)

What is reported

  • Backscattering profile, every scan, with elapsed time
  • Foaming Capacity (%): (foam top height − vial floor) / pre-foam liquid column x 100. At 100% the top of the foam reached the level of the un-foamed liquid, so there was no net expansion; values above 100% measure how far foaming raised the column.
  • Foaming Stability (%): foam height on the last scan (30 min) / foam height on the first scan x 100
  • Foam Drainage (%): (foam height on the first scan − foam height on the last scan) / foam height on the first scan x 100. This is the share of the foam column lost over 30 minutes, the complement of foam stability, and the two sum to 100%. It is reported under this name for continuity with earlier data; it measures loss of foam height, not the volume of liquid draining out of the foam.
  • Sediment layer thickness (mm), final and maximum, and its growth rate while forming (mm/min)
  • Clarification front velocity (mm/min)
  • Change in backscattering at the vial floor and across the liquid (%), last scan minus first

Heights are measured from the vial floor, taken as 5 mm above the instrument's optical zero, since readings close to the glass are unreliable. The pre-foam liquid column is the height of 10 mL of liquid above that floor, read from a reference vial filled and scanned without foaming (about 20 mm), and is applied to every run.

Full principle

Capacity favors proteins that reach the air-water interface quickly and rearrange there. Because the air-water interface is strongly hydrophobic, a protein generally must unfold to some degree on adsorption to form a coherent film, so conformational flexibility rather than native structure tends to be the useful property here. Compact, disulfide-stabilized globulins often adsorb slowly and foam poorly, while flexible or partially denatured proteins adsorb and spread more readily. Stability, by contrast, depends on the mechanical properties of the film once formed, and is favored by viscoelastic interfacial layers reinforced by hydrophobic contacts or disulfide exchange after adsorption. The two properties often trade against each other, which is why both are measured.

Dissolved protein adsorbs most readily, so foaming behavior largely reflects the soluble fraction rather than the total protein loaded, though colloidal protein particles in these unclarified dispersions can also influence bubble formation and stability. Results are therefore best interpreted alongside solubility determined on the same material and at the same pH. Foaming values are also strongly method-dependent, since gas incorporation depends on the geometry, speed, and duration of aeration, and are comparable only within this protocol.

The Turbiscan scans the full height of the vial every two minutes for 30 minutes, recording backscattering and transmission at each height, and every scan is kept with its elapsed time, taken from the instrument's own time stamp. Capacity is read from the first scan, taken within 60 s of foaming. Stability compares the foam column on the last scan with the first. Because the whole series is kept, the analysis also follows the liquid under the foam: whether material settles onto the vial floor, and whether the liquid clears from the top down as it does. Stability and the settling readouts describe early destabilization over 30 minutes, and should not be read as a measure of foam lifetime.

How the profile is read

Foam scatters far more light than the liquid below it or the air above it, so on each scan the foam appears as a band of high backscattering. Its edges are placed where backscattering crosses a line drawn 11% of the way from the background level up to the peak of the band, interpolated between measurement points. The 11% level was calibrated against a reference vial holding a foam column of known height (7.5 mm), which it reads to within 0.4%.

Two situations need more than that line. A turbid liquid, as in some pumpkin and hemp dispersions, can scatter more strongly than the line itself, so the lower edge would run through the liquid to the vial floor. When an edge reaches the floor or the top of the vial, it is re-measured against the level of the liquid, or the air, immediately next to the foam rather than the overall background; runs whose edges fall clear of the floor and top are read exactly as before. Separately, material settling out of a turbid dispersion can come to scatter more than the foam. The foam is therefore taken as the uppermost distinct band in the vial rather than the brightest, and on the last scan it is sought within the zone the foam occupied on the first scan, since the foam at the end is what remains of the foam at the start.

Settling is read by comparing every scan with the first. Material collecting on the vial floor raises backscattering there. The sediment layer is the region above the floor where that rise is at least half the largest rise seen at the floor during the run, and a run is counted as settling only when the rise at the floor exceeds 5 backscattering percentage points, which is above scan-to-scan noise. As particles leave the liquid above the layer, the liquid clears from the top down. The clarification front is the lowest point in the liquid that has cleared by at least half the deepest clearing seen just under the foam, and its velocity is the rate at which that point descends; a liquid that clears evenly throughout has no front. The signed changes at the floor and across the liquid distinguish settling, which raises the floor and lowers the liquid, from other kinds of change. The Turbiscan Stability Index cannot draw that distinction, because it sums change without regard to direction.

Foam values are taken from the first and last scans, and the reading is also checked on every scan in between. A run is marked low confidence when a middle or final scan shows no foam, or when the lower edge of the foam steps down by more than 1 mm from one scan to the next. The confidence mark is stored alongside the foam values and does not alter them.

Low confidence arises most often in thin foams. A foam only a few millimeters tall stands out only slightly from the liquid below it and the air above it, so the boundary sits close to the limit of what the profile can resolve, and three patterns follow from that. In the first, the foam is not found on one or more middle scans and is found again afterwards at a consistent height. A foam does not disappear and re-form, so this is the reading losing a faint band, and the first and last scans are unaffected by it. In the second, the lower edge drops abruptly partway through the run. The lower edge normally rises as the foam drains, and a foam cannot grow back, so a downward step or an apparent regrowth marks a scan on which the boundary was placed on the wrong feature. In the third, no foam is found on the final scan. The foam may have collapsed completely, or thinned below what the profile can detect, and the reading cannot distinguish the two; in that case the stability value depends on a final foam height that the full series does not support. Because it arises mainly in thin foams, low confidence is itself informative about the material: it usually marks a protein that forms little foam under these conditions, and whose foam values are consequently less certain than those of a material that foams well.

Equipment
  • Turbiscan Tower (Formulaction)
  • Rotor-stator homogenizer (PRO Scientific PRO25D, 10 mm probe)
  • Turbiscan vials
  • Homogenizer stand
Consumables
  • 0.1 M phosphate buffer (pH 7.0)
  • 1 M NaOH/ 1 M HCl
Procedure
  1. Dilute samples. Standardize samples to 1% protein (w/w) at pH 7.0 ± 0.1 with 0.1 M phosphate buffer (pH 7.0). Adjust with 1 M NaOH or 1 M HCl. Stir dispersions for 30 min at room temperature, and readjust pH as needed.
  2. Turbiscan setup. Set Temperature Control to 22 °C, and allow the instrument to equilibrate before testing. In the Towersoft software, set up the sample name and analysis sequence for Foaming: Starts immediately, 16 scans every 2 min (30 min). The realized interval can be longer than the set one, so a run may record fewer scans; each scan's elapsed time is taken from the instrument's own time stamp rather than assumed from the schedule.
  3. Foam dispersion. In a Turbiscan vial, add 10 mL 1% protein dispersion. Record the initial liquid height. Use a stand to position the vial so the liquid surface sits in the middle of the bottom probe holes. Homogenize the sample for 2 min at 10,000 rpm, keeping the vial stationary with the homogenizer probe in the middle. Move the vial as needed if the surface level changes with foaming, so that the probe holes stay at the liquid surface. This adjustment is made by eye and is the largest operator-dependent variable in the method.
  4. Load sample. Immediately cap the vial, and place it into the Turbiscan tower for measurement. The measurement will begin when the door closes. Load within 60 s of the end of homogenization, since the first scan establishes the initial foam height. Repeat in triplicate.
  5. Sample removal. When samples are done running, ensure the green bars are shown before opening the doors and removing the vials. When finished, turn off the Temperature Control.
  6. Data quality check. Check the auto-detected meniscus, and adjust height as needed. The bottom of the vial should be around 5mm and the meniscus should be at the top inflection point of backscattering from the first timepoint.
References
  1. Schwenke, K. D., Rauschal, E. J., & Robowsky, K. D. (1983). Functional properties of plant proteins Part IV. Foaming properties of modified proteins from faba beans. Food / Nahrung, 27(4), 335–350. https://doi.org/10.1002/food.19830270407
  2. Ozkan, G., Tataroglu, P., Gulec, S., & Capanoglu, E. (2024). Modification of pea protein isolates by high-intensity ultrasonication: Functional, structural and nutritional properties. Food Chemistry Advances, 5, 100793. https://doi.org/10.1016/j.focha.2024.100793
  3. Halm, J., Nyhan, L., Zannini, E., & Arendt, E. K. (2026). Plant-based proteins as functional egg replacers in pound cake: A comparative study of legume and oat ingredients. Current Research in Food Science, 12, 101373. https://doi.org/10.1016/j.crfs.2026.101373
In the PDF, p. 46

Gel Strength

A protein gel is a continuous three-dimensional network that immobilizes the surrounding liquid. Heating unfolds dispersed protein and exposes buried hydrophobic surfaces and free thiol groups, and the unfolded molecules aggregate into a network through hydrophobic association and thiol-disulfide exchange. Cooling reinforces that network, largely through hydrogen bonding, which strengthens as temperature falls. The relative contribution of each interaction is protein- and condition-dependent rather than a fixed sequence. Gel strength, the resistance of the set gel to a defined deformation, is the headline functional property for any application that relies on a gel matrix to hold structure, including custards, surimi-style products, and meat analogues.

Instrument

Stable Micro Systems TA.XTPlusC Texture Analyzer

What is reported

  • Hardness (g)
  • Cohesiveness
  • Springiness
  • Gel observations
Full principle

Gelation requires a minimum protein concentration, below which no continuous network forms, and above that threshold strength rises steeply with concentration, commonly with an exponent between two and four depending on the protein and on how far above the critical concentration the measurement sits. Comparisons are therefore valid only at matched protein concentration, and errors in the nitrogen-to-protein conversion factor propagate into gel strength amplified by that exponent. Network architecture also depends on pH and ionic strength: near the isoelectric point, or at high ionic strength, charge screening produces coarse particulate gels that are opaque and prone to syneresis, while conditions far from the isoelectric point favor fine-stranded networks that are translucent and hold water better. Since residual salt varies by extraction route, gel results are read alongside ash content from the same material. Gels are set at pH 7.0 while DSC thermograms are recorded at extract pH, so the two are related but not directly comparable.

Texture Profile Analysis (TPA) is the standard instrumental measurement. The sample is compressed twice in succession by a probe at fixed speed and depth, an approach originally conceived as an imitation of two consecutive chews. The resulting force-time curve yields hardness, cohesiveness, springiness, chewiness, and resilience. TPA is an empirical test rather than a fundamental rheological measurement: the values depend on sample geometry, probe diameter, compression ratio, and crosshead speed, so results are comparable only between samples run under identical conditions. Gel strength is reported here as TPA hardness, the peak force during the first compression.

Gels are cast to 20 mm depth and indented with a 12 mm flat probe to 20% strain. These conditions are constant across the full dataset. Because the probe is smaller than the sample container, this is a local indentation rather than the bulk compression of classical TPA. The surrounding gel is unconstrained and the second compression acts on a surface potentially already disrupted by the first, so hardness as the peak first-compression force remains a sound relative measure, while cohesiveness and springiness derived from the second cycle should be read as indices specific to this geometry and not as classical TPA parameters. Relative comparison within this dataset is the intended use, and comparison outside it should be made with care.

Where a sample does not form a self-supporting gel, the instrument will still return a peak force from the viscous dispersion. Such samples are recorded as non-gelling in the observations field and their hardness value flagged, so that they are distinguishable in the dataset from weak but continuous gels.

Equipment
  • 2 oz stainless steel cups
  • Stable Micro Systems TA.XTPlusC Texture Analyzer
  • 12 mm diameter acrylic cylindrical probe (TA-10)
  • Sample platform (TA-90)
  • 2 kg weight
  • Water bath
Consumables
  • DI water
  • 1 M NaOH/ 1 M HCl
Procedure
  1. Dilute samples. Standardize samples to 10% protein (w/w) at pH 7.0 ± 0.1 in DI water. Adjust with 1 M NaOH or 1 M HCl. Stir dispersions for 30 min at room temperature, and readjust pH as needed. Record the actual protein % achieved. A sample run below 10% protein is flagged as informational.
  2. Gel samples. Pour 30 g of dispersion into a metal cup (~20mm sample height). Create in triplicate. Ensure the liquid surface is flat. Cover with foil and heat in a 90 °C water bath for 30 min. Cool at room temperature for at least 1 hour.
  3. Texture Analyzer set up. On the Texture Analyzer software, configure TPA settings: 1mm/s test speed, 20% strain target, 5 g trigger force, 1 sec interval between compressions, 5mm/s pre/post-test speed. Calibrate force with 2 kg weight, and calibrate probe height on the sample platform.
  4. Measure TPA. Place the gel under the probe so it presses near the center. Test each gel twice in different locations on the surface. Position both tests at least one probe diameter from the cup wall and from each other. The second indentation is a second reading on the same gel rather than an independent replicate. Wipe the probe between runs, and ensure the measurement starts when the probe contacts the gel.
  5. Data quality check. Confirm 2 consecutive peaks starting at 0 seconds.
References
  1. Friedman, H. H., Whitney, J. E., & Szczesniak, A. S. (1963). The Texturometer-A New Instrument for Objective Texture Measurement. Journal of Food Science, 28(4), 390–396. https://doi.org/10.1111/j.1365-2621.1963.tb00216.x
  2. Joyner, H. S. (2024). Rheological Principles for Food Analysis. In B. P. Ismail (Ed.), Nielsen's Food Analysis (6th ed., Ch. 29). Springer. https://doi.org/10.1007/978-3-031-50643-7_29
  3. Boyle, C., Hansen, L., Hinnenkamp, C., & Ismail, B. P. (2018). Emerging Camelina Protein: Extraction, Modification, and Structural/Functional Characterization. Journal of the American Oil Chemists' Society, 95(8), 1049–1062. https://doi.org/10.1002/aocs.12045
In the PDF, p. 50

Appendix A (factor decision history) and Appendix B (routine measurements) are in the full compendium.