Corollary

Research

  • Papers

Library

  • Catalog

Account

  • Jobs
  • Billing
SettingsResearch demo · not for clinical or commercial use
Corollary
  1. Catalog
  2. Protein Design
  3. SaProt

Protein Design · Su et al., Westlake University

SaProt

Structure-aware protein LM using combined sequence + structure tokens.

Commercial use permittedThe MIT licence allows use in commercial research and products.MIT
Availability
On demand
Typical latency
2.5s
Credits per run
~14
Compute tier
A · Fast

1–30s

Runs on On-demand GPU

Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.

Run this model

SaProt

Single-letter amino acid code. Whitespace and FASTA headers are stripped automatically.

Optional, but it is the reason to use this model rather than ESM2.

Comma-separated, e.g. A123V, W45F. Leave empty for an embedding.

Needs protein sequence

About this model

SaProt tokenizes structure alongside sequence with Foldseek's 3Di alphabet, giving representations that beat ESM2 on many functional prediction tasks. Worth the extra structure input whenever you have one.

Standardized I/O contract

Every model in the Hub speaks the same contract, which is what lets the Router and the agent call any of them without special-casing.

Inputs

  • sequencesequence · required

    Protein sequence

  • pdbtextarea

    Structure (PDB or entry code)

  • mutationstext

    Mutations to score

Outputs

  • embeddingvector

    Structure-aware representation.

Specification

Hardware
1× A100 16GB
GPU memory
16 GB
Version
1.0
Licence
MIT
Backend
On demandOn-demand GPU
MCP server
mcp-protein-server

Tasks

sequence embeddingvariant effect

Source

  • Model card
embeddingstructure-awaremit-license