BioAnalix makes proteins and enzymes hold up under the conditions that would otherwise destroy them — heat, storage, and process stress. We run this as a repeatable engineering service, not a one-off research project.

The loop

Models propose. Screening measures. Measurements correct the models. Each turn narrows the search, so later rounds test fewer variants and find more of them useful.

DESIGNmodels propose variants BUILDlibrary constructed TESThigh-throughput screen LEARNmodels retrained EACH CYCLE narrows the search

The model architecture, screening design, and selection conditions are proprietary. In customer engagements they are covered under NDA.

What successive rounds look like

The useful signal is not how many variants survive a single screen, but how quickly the population concentrates. An enrichment curve makes that visible: the steeper the climb across rounds, the fewer designs need to reach the bench.

0%25%50%75%100% Round 1Round 2Round 3Round 4Round 5 Illustrative shape of enrichment across rounds — not measured programme data Share of variants retaining function

The BioAnalix Workbench

The Workbench is the analysis environment behind the loop. It takes sequence and measured stability data, fits predictive models against them, and ranks proposed designs before any of them reach the bench.

Engagements include the outputs and the reasoning: ranked variant lists with predicted stability, enrichment curves showing how successive rounds concentrate useful variants, and position-level maps showing where a sequence tolerates change and where it does not. The underlying feature set and model configuration stay in-house.

Why this is a hard problem

Stabilising mutations do not simply add up. A substitution that helps on its own can hurt in combination with another, and predicting those interactions from structure alone remains unsolved in the general case. Our approach does not assume the model is right. It treats every round of measurement as a correction to the model, which is why the loop converges on designs that pure prediction misses.

This is not a new interest for us. Our co-founder Dr. Hassan Pezeshgi Modarres published on enzyme thermostability engineering and built ProtDataTherm, a database of the sequence and structural features that separate thermostable proteins from their mesophilic counterparts, before BioAnalix existed.

Track record

  • Incorporated in 2021 and operating for five years
  • Three consecutive years of accepted SR&ED claims
  • Recurring revenue from university and hospital research customers in Canada and the United States
  • Affiliated with H2i, Temerty Faculty of Medicine, University of Toronto
  • Based at 100 King St West, Toronto

Talk to us

For scoping conversations and technical detail under NDA: info@bioanalix.ca