Short label for this ParaSurf run. It appears in your dashboard and job history. It does not change how the model runs.

What do these mean? ParaSurf can load weights trained on different antibody datasets. Pick the one that matches your use case.

• PECAN Dataset — General-purpose antibody–antigen complexes curated from structure databases. Good default when you are unsure which to pick.
• Paragraph Expanded — Broader training derived from the Paragraph paratope dataset. A solid choice for overall paratope surface prediction across many antibody types.
• Paragraph Expanded (Heavy Chains Only) — Training focused on heavy chains. Use this when your analysis centers on VH or when the light chain is incomplete or low quality.
• Paragraph Expanded (Light Chains Only) — Training focused on light chains. Use this when VL is the primary area of interest or when the heavy chain is incomplete or noisy.
• MIPE Dataset (5-fold CV) — Trained and validated with five-fold cross-validation on the MIPE set. Choose this for validation-balanced behavior in benchmarks or publications.

You can rerun the same structure with different weights to compare surface maps and pick what best matches your experimental expectations.

What is this? We turn your protein into a 3D surface made of small facets called a mesh. The mesh density controls how many facets we use.

Speed vs detail: 0.1 is faster. Higher values create more facets, take longer, and can capture more fine-grained surface details.

Default: 0.3. Increase only if you need more detailed features and can allow longer runtime.

Valid values: 0.1, 0.2, …, 1.0