Plant Protein Replacement at Scale
Abstract
The plant based meat company needed to replace a key protein in a live product, and optimise a costly ingredient in a product in development, both decisions carrying high commercial risk, both needed in weeks. Using Liquid state intelligence, Apoha identified two viable protein substitutes from 15 candidates, one of which is now in the product on the market, and found that as little as 5ľ15% of a high-cost functional ingredient matched the performance of pure inclusion, replacing months of trial-and-error with confident, evidence-based decisions.
In Collaboration with
How to read this
Investigation 1
Protein Substitution: Screening 15 candidate proteins to identify a replacement for a core ingredient in a live product already on supermarket shelves.
Investigation 2
Cost Optimisation: Identifying the minimum effective inclusion level of a high-cost functional ingredient in a product in development.
About Liquid State Intelligence
A new class of data – molecular behaviour
Liquid State Intelligence is Apoha's measurement platform for capturing how molecules behave under real-world stresses, where performance actually matters. Most tools measure one isolated property at a time. Liquid State Intelligence delivers a small sample onto a specially engineered liquid surface and reads the waves that result, delivering a behavioural picture of that molecule in a single measurement, from as little as 10µg of material.
Developability failures happen late because early-stage tools don't capture the full picture of how a molecule will behave under real conditions. Liquid State Intelligence gives R&D teams a single high-dimensional readout sensitive to aggregation tendency, hydrophobicity, stability, and more, from one measurement, as early as hit identification. Apoha's platform eliminates candidates destined to fail with high precision, so the molecules that continue down the development pipeline are the ones most likely to reach patients.
Applied to monoclonal antibodies, each droplet impact subjects the molecules to a transient, high-shear coalescence event that simultaneously imposes interfacial expansion, Marangoni-driven surface compression, axial elongational flow, and a sharp chemical-potential gradient, a single mechanically and chemically resolved probe of behaviours linked to aggregation, surface activity, viscosity and self-association.
Customer Background
The plant based meat company is a plant-based food company that creates hyper-realistic meat alternatives for meat-eaters.
The plant based meat company faces formulation decisions both in active development and in live products already on shelf. In plant-based food, these decisions carry high stakes, ingredient changes can affect product quality and brand reputation, and plant-based ingredients vary dramatically in cost, making it essential to get formulation right to protect margins. These decisions often need to be made in weeks, not months.
Plant protein supplier specifications typically cover purity and pH, useful as a starting point, but they say little about functional behaviour. Viscosity, solubility, and gelling can be measured independently, but how a protein performs within a complex formulation matrix, and under real manufacturing conditions, is difficult to predict. Plant proteins respond differently to different environments, and their behaviour in a product is nonlinear, meaning small changes in formulation or process can produce disproportionate effects on the final result.á
The challenge
Protein Substitution
The plant based meat company needed to replace a key ingredient in a product already on supermarket shelves, with only weeks to make the decision. Getting it wrong meant risking product quality, customer trust, and production disruption.
The product was in active production and any failure would have immediate commercial consequences, either a batch that failed to be released, or a quality drop visible to consumers. Conventional ingredient data gave a useful starting point but not enough confidence to act on alone. With 15 candidate proteins to evaluate and no time for extended physical trials, they needed a faster, more confident way to identify the closest functional match.
How characterisation was applied
The plant based meat company had 15 candidate proteins to choose from but no fast way to know which would actually work in their formulation. Apoha compared all 15 against the original protein, then ran deeper tests on the top 4 within the actual product formulation, giving the plant based meat company a evidence-based shortlist in weeks. Apoha used a two-stage approach to provide this. Stage 1 benchmarked all 15 candidates against the gold-standard reference protein to identify the closest behavioural matches. Stage 2 tested the top 4 candidates within a partial formulation matrix containing the key functional ingredients, capturing how each protein behaves in the presence of the components that most influence product performance.
As a long-term trial was not an option, the plant based meat company had conducted conventional physicochemical testing on the candidates but needed greater confidence in choosing a protein replacement, due to the commercial risk if product quality dropped.
Each candidate protein was measured at controlled concentration, generating surface pressure data and a high-dimensional fingerprint per sample. Surface pressure measurements provided a first-pass comparison of interfacial adsorption behaviour across all 15 candidates. PCA was then applied to the fingerprint data, enabling visual clustering and quantitative similarity scoring relative to the gold-standard reference. Surface pressure was captured as a separate readout.áThe top 4 candidates, which also matched with internal characterisation, were re-measured within the actual base formulation matrix, capturing how each protein behaves in the presence of the other formulation components, where interaction effects determine real functional performance.
Results, outcomes and key data
Some proteins looked similar on the surface pressure measurement, but deeper analysis revealed which ones truly matched the target behaviour. Liquid state intelligence identified two viable alternatives. The plant based meat company then considered the cost and availability of each, and the candidate that best matched the target behaviour also proved the most viable commercially. It is now in the product on the market.
Figure 1 shows surface pressure measurements across all 15 candidates, with a dotted line indicating the gold-standard reference value. Several candidates sit close to the reference level, providing a useful first-pass filter, but surface pressure alone was not sufficient to shortlist with confidence. Figure 2 shows PCA of the full fingerprint data provided the additional resolution needed, identifying two candidates whose combined behavioural profile most closely matched the gold standard, which would have been difficult to distinguish by surface pressure alone.
Plant proteins are known to be highly variable in both composition and behaviour, even within the same protein type. Figure 2 shows this variability clearly, with the four candidates displaying a range of profiles in PCA space. Despite this, a clear overlap in behaviour between S1 and the gold standard S2 is visible, with S3 showing partial overlap as a secondary candidate.
L3/4 Combining the PCA results with the surface pressure measurements, two candidates emerged as the closest matches to the gold standard. These results were also consistent with the physicochemical data the plant based food company had measured independently. Both were taken forward for physical trials. One candidate was selected, successfully trialled at scale, and incorporated into the recipe, with the product now on the market.
Plant proteins are known to be highly variable in both composition and behaviour, even within the same protein type. Figure 2 shows this variability clearly, with the four candidates displaying a range of profiles in PCA space. Despite this, a clear overlap in behaviour between S1 and the gold standard S2 is visible, with S3 showing partial overlap as a secondary candidate.
L3/4 Combining the PCA results with the surface pressure measurements, two candidates emerged as the closest matches to the gold standard. These results were also consistent with the physicochemical data the plant based food company had measured independently. Both were taken forward for physical trials. One candidate was selected, successfully trialled at scale, and incorporated into the recipe, with the product now on the market.
How characterisation was applied
The plant based meat company needed to find the right amount of a high-cost ingredient to include in a new product, too much would destroy margins, too little could compromise quality. Apoha tested the formulation across a range of inclusion levels to find the sweet spot where a small amount delivered the same performance as much more.
Rather than months of iterative trial-and-error, Apoha used a single structured study to map how formulation performance changed across inclusion levels of the functional ingredient within the base matrix. This identified the minimum threshold of the high-performing but high cost ingredient needed to maintain product quality, enabling the plant based meat company to make a confident, evidence-based decision in weeks rather than months.
Liquid State Intelligence was used to characterise formulation behaviour across a range of inclusion levels of the high-functionality ingredient within the base matrix. Rather than measuring the ingredient in isolation, the analysis focused on interaction effects between the functional ingredient and the broader formulation, capturing the emergent, nonlinear behaviour that governs real product performance. PCA of the high-dimensional fingerprint data was used to identify the inclusion regimes where behaviour shifted most significantly, mapping the inclusion-response relationship and identifying where lower inclusion levels produced behaviour overlapping with the high-inclusion reference. The Liquid State Intelligence was able to compare complex, highly similar materials.
.png)
.png)
Results, outcomes and key data
Formulations with higher proportions of the high-functionality protein cluster in the upper region of PCA space, while the 0:100 formulation sits distinctly separate. Notably, the 25:75 ratio shows the greatest spread from the 100:0 reference, suggesting that at this inclusion level the high-functionality protein may behave differently in the presence of the lower-functionality protein, though the mechanism is not fully understood and this observation requires further investigation. Formulations at 5:95, 10:90, and 15:85 show the closest behavioural overlap with the 100:0 reference, indicating that as little as 5ľ15% high-functionality protein achieves comparable performance to the pure high-functionality formulation.
This identified a clear sweet spot where a low inclusion level achieved performance overlapping with much higher-cost formulations. The plant based meat company could therefore use the expensive ingredient strategically rather than maximally, preserving product specifications while significantly improving cost efficiency. Rather than months of trial-and-error or an educated guess that could have introduced significant doubt into the project, the decision was made with confidence and without the time needed for a physical trial.
Formulations with higher proportions of the high-functionality protein cluster in the upper region of PCA space, while the 0:100 formulation sits distinctly separate. Notably, the 25:75 ratio shows the greatest spread from the 100:0 reference, suggesting that at this inclusion level the high-functionality protein may behave differently in the presence of the lower-functionality protein, though the mechanism is not fully understood and this observation requires further investigation. Formulations at 5:95, 10:90, and 15:85 show the closest behavioural overlap with the 100:0 reference, indicating that as little as 5ľ15% high-functionality protein achieves comparable performance to the pure high-functionality formulation.
This identified a clear sweet spot where a low inclusion level achieved performance overlapping with much higher-cost formulations. The plant based meat company could therefore use the expensive ingredient strategically rather than maximally, preserving product specifications while significantly improving cost efficiency. Rather than months of trial-and-error or an educated guess that could have introduced significant doubt into the project, the decision was made with confidence and without the time needed for a physical trial.
.png)
Next steps
Following this collaboration, Apoha has continued to expand its work with plant-based proteins to working with different companies with different priorities. Further case studies are in development.