StrataBio
StrataBio combines high-throughput protein stability experiments with AI/ML to help protein-engineering teams prioritize which variants are worth testing. The platform centralizes experimental data scattered across assay files, spreadsheets, and lab instruments to close the loop between computational design and experimental performance. Initial models on small domain sequences achieve 0.687 kcal/mol MAE and 0.858 Pearson correlation for ddG at room temperature.
- Biotechnology
- Artificial Intelligence
- Drug Discovery & Therapeutics
- Data & Analytics
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Protein-engineering teams generate valuable experimental data across every design-build-test-learn campaign, but results are often scattered across assay files, spreadsheets, laboratory instruments, and different research groups. This fragmentation makes it difficult to systematically learn which sequence, structural, and experimental factors are associated with improved protein stability, and computational predictions of protein performance frequently fail to match experimental reality.
Solution
StrataBio combines high-throughput protein stability experiments with AI/ML to assist protein-engineering teams in prioritizing the variants most worth testing. The platform is being built to learn from experimental data already generated across expression, folding, thermal stability, retained folding after stress, and related durability phenotypes. By centralizing and analyzing these datasets, StrataBio aims to close the loop between computational design and experimental performance, making every experiment more informative. Initial model results on small domain (40–80 amino acid) sequences show an MAE of 0.687 kcal/mol and Pearson correlation of 0.858 for ddG at room temperature.
Target Audience
Protein-engineering teams in biotech and pharmaceutical companies that need to determine which variants are stable enough to justify further development, alongside their existing tools for discovering active proteins and optimizing catalytic function.
Features
- AI/ML models trained on high-throughput experimental protein stability data (ddG) rather than purely computational predictions
- Centralized data aggregation across assay files, spreadsheets, campaigns, laboratory instruments, and research groups
- Initial focus on expression, folding, thermal stability, retained folding after stress, relative stability, and durability phenotypes
- Expandable to process-relevant conditions such as solvent tolerance, pH extremes, and salt tolerance as datasets grow
- Benchmarked accuracy: 0.687 kcal/mol MAE, 1.029 kcal/mol RMSE, 0.858 Pearson, 0.861 Spearman on small domain sequences