Generates and prioritizes protein sequences for design campaigns, including binder discovery.

DSM

DSM is Synthyra's masked-diffusion protein generation direction. It is built for exploring protein sequence space and ranking candidate designs before experimental follow-up.

What It Does

DSM supports:

• Generating protein sequences from masked sequence context.
• Exploring variants around known templates.
• Producing candidate binders for design campaigns.
• Pairing generation with Atlas scoring, annotation, and structure checks.
• Prioritizing candidates before synthesis or wet-lab screening.

Why It Matters

Protein design is a search problem. The number of possible sequences is enormous, while the number a team can synthesize and test is small.

Masked diffusion lets DSM use global sequence context while filling in missing residues. That is a useful fit for proteins because residues far apart in sequence can become neighbors after folding.

The DSM research also moved beyond ranking alone. Repo-local content notes experimental validation on EGFR and PD-L1 candidates, including real expression and binding signals and a sub-nanomolar EGFR binder.

Intended Use

Use DSM for early design exploration when a team needs candidate sequences to rank, inspect, and validate. It is most useful alongside Atlas, structure prediction, annotation, developability oracles, and wet-lab follow-up.

Limitations

DSM does not guarantee expression, folding, binding, safety, or function. Generated proteins can fail for reasons outside the model's view. Treat designs as hypotheses that require responsible review and experimental validation.