Usage

You can predict the melting temperature (Tm) of a nanobody (VHH) sequence two ways: a simple web UI, or a REST API for programmatic use.

Web UI

Open https://diffuse-bio--proxytm.modal.run/ in a browser, paste a single-letter VHH amino acid sequence into the input, and press Enter. The page shows the predicted Tm (°C) and the std across the model ensemble.

The UI is prefilled with an example sequence so you can hit Enter to see the format.

Rest API

Example request:

GET <https://diffuse-bio--tm.modal.run/?seq=><AMINO_ACID_SEQUENCE>

Example:

curl '<https://diffuse-bio--tm.modal.run/?seq=EVQLLESGGGLVQPGGSLRLSCAASGMDIDYQAMSWFRQAPGKEREFVSCISATGTVTYYADSVKGRFTISRDNSKNTLYLQMNSLKPEDTAVYYCGFSPDCFGERGLYDVSDSRWGQGTQVTVSS>'

Response:

{"pred_tm": 67.42, "std": 1.83}

• Request takes a few seconds and returns a predicted Tm and standard deviation across the model ensemble
• Right now, only VHH input sequences are supported
• For large batch requests, get in touch
• pred_tm is the mean prediction across an ensemble of 5 fine-tuned AntiBERTa2 models; std is the disagreement between ensemble members (a rough confidence signal — high std means the models disagree).
• Request takes a few seconds (longer on cold start) and returns a single JSON object.
• Only VHH input sequences are supported. Sequence must be amino-acid letters only (no X), max 1000 residues.
• Rate limit: 5 requests/minute per client.
• For large batch requests, info@diffuse.bio.