Structure prediction · Protein Design

ESMFold 2

Fold a protein from its sequence alone, with no MSA or template search.

Writes a FASTA record and invokes the esm-fold CLI from fair-esm. Predictions are single-sequence, so no alignment or template database is consulted.

Documentation ↗ Home page ↗ Source code ↗

Input

Configure a small job

POST

Job name

Protein sequence Join chains of a complex with a colon to fold them as one record.

Recycles More recycles refine the prediction at a proportional runtime cost.

Max tokens per batch

Attention chunk size 0 disables chunking; lower values trade speed for less memory.

CPU only

Offload parameters to CPU Keeps large sequences within GPU memory at the cost of speed.

Ready

Response

{
  "status": "Ready for a request"
}