Prediction of TCR-p complexes.

[Alexasparis]

I am using HelixFold3 for modeling TCR-pMHC complexes, and it works when modeling protein chains. However, I encounter errors when I include the peptide in the complex.

Could you clarify whether HelixFold3 supports peptide modeling? If so, should the peptide be configured as a protein or as a ligand in the input?

Thank you for your assistance.

[xueeinstein]

@Alexasparis HelixFold3 supports to model peptide as protein chain ligand or ligand chain. If use ligand chain, you should use SMILES to represent the peptide.

Example: model peptide as ligand chain B:

{
    "entities": [
        {
            "type": "protein",
            "sequence": "ADVPAGVQLADKQTLVRNNGSEVQSLDPHKIE.....",
            "count": 1,
            "chain_id": "A"
        },
        {
            "type": "ligand",
            "smiles": "CC(C)C[C@H](NC(=O)[C@@H](N)CCCCN)C(=O)N[C@@H](CCCCN)C(=O)O",
            "count": 1,
            "chain_id": "B"
        }
    ]
}

Example: model peptide as short protein chain B:

{
    "entities": [
        {
            "type": "protein",
            "sequence": "ADVPAGVQLADKQTLVRNNGSEVQSLDPHKIE.....",
            "count": 1,
            "chain_id": "A"
        },
        {
            "type": "protein",
            "sequence": "KLK",
            "count": 1,
            "chain_id": "B"
        }
    ]
}

[Alexasparis]

The problem is that when modeled as a protein, it raises the exception. Template too short. Length: 9 and uniptor_hits.sto is not generated for the peptide chain.