Challenge Provider: RLA lab, Imperial College London (Oliver Konzock)
Team: Miranda Shou, Alexia Horea, Ricardo Valencia:, Ziyi Cheng
Mentors: Technical: Oliver Konzock, Geoff Baldwin, Pitch: Tatia Codreanu
The Challenge
Alternative proteins from plants, fungi, algae, and microbes have immense potential to support a sustainable and resilient food system. However, many suffer from incomplete essential amino acid (EAA) profiles and poor digestibility, limiting their nutritional performance compared to animal proteins.
The challenge was to develop a computational and AI-assisted workflow capable of screening vast protein databases to identify highly digestible, EAA-rich candidates. The goal was to build a scalable bioinformatics pipeline that could predict digestibility and nutritional quality directly from sequence and structure data, enabling more targeted discovery of superior protein sources for food and feed.
The Approach
The team developed a prototype analytical platform, BiteScore, that integrates bioinformatics, machine learning, and protein structure prediction into a unified pipeline. The system combines data from AlphaFold, UniProt, and NCBI to extract sequence and structure features relevant to digestion, while applying predictive algorithms benchmarked against FAO/WHO protein quality standards (PDCAAS and DIAAS). Protease recognition site analysis was used to estimate cleavage accessibility and overall digestion efficiency, while comparative scoring and visualization enabled ranking of multiple proteins according to their predicted digestibility and nutritional balance.
To make the results interpretable for end users, LLM-assisted summaries provided contextual insights and recommendations for further experimental validation. By benchmarking the workflow against well-characterized proteins such as milk and egg, the team ensured both biological plausibility and predictive robustness of the model.
The Outcome
The resulting BiteScore App is a functional prototype capable of predicting and visualizing protein digestibility in minutes. It enables standardized, reproducible, and ethically sound assessment of protein quality, reducing dependence on labor-intensive in vivo experiments. By automating digestibility prediction and integrating seamlessly with proteomic workflows, BiteScore has the potential to accelerate R&D pipelines across food technology, nutritional science, and synthetic biology. Ultimately, this tool lays the groundwork for data-driven protein design, guiding the development of more nutritious and sustainable ingredients for the future food system.
Impact & Next Steps
BiteScore accelerates R&D by providing rapid, in silico prediction of protein quality, reducing the time and cost for screening new alternative protein sources. The tool minimizes reliance on resource-intensive in vivo trials, promoting more ethical and sustainable nutritional assessment methods. BiteScore moves protein development to a data-driven approach, allowing researchers to design or select proteins with optimized essential amino acid (EAA) profiles and enhanced protease cleavage sites. Finally, the platform offers a standardized, reproducible metric for protein quality (predictive PDCAAS and DIAAS equivalents) to aid regulatory bodies and consumers in making informed decisions.
The immediate next steps for BiteScore involve Experimental Validation through partnerships with research groups, using in vitro and in vivo models to refine the predictive algorithms. To enhance its utility, the platform will undergo Feature Expansion to include flagging of detrimental factors from the protein sequence, including allergenicity screening. A key development is the creation of a public API Development to allow seamless integration of BiteScore into existing industry and academic proteomic analysis workflows. Finally, the team plans to explore Open-Source Contribution through paper publication and potentially releasing core methodology to benefit the broader food and protein discovery communities.