Skip to content

module adabmDCA.api.contacts

High-level contact-prediction operations.

function predict_contacts

predict_contacts(
    model: 'DCAModel | str | Path | None' = None,
    fasta_path: 'AlignmentInput | None' = None,
    alphabet: 'str | None' = None,
    pseudocount: 'float' = 0.5,
    device: 'str' = 'auto',
    dtype: 'str' = 'float32'
) → ContactMapResult

Predict residue contacts from a DCA model or directly from an alignment.

With model, scores are the average-product-corrected (APC) Frobenius norms of the zero-sum-gauge couplings, excluding the gap state. With fasta_path, a mean-field DCA model is inferred from the reweighted alignment and scored the same way. Larger scores mean likelier contacts.

Args:

  • model: A DCAModel, or a path to a parameter file or PTT archive.
  • fasta_path: Aligned sequences (path, Alignment or sequence strings) for mean-field prediction. Provide exactly one of model and fasta_path.
  • alphabet: "protein", "rna", "dna" or an ordered custom token string, for the alignment or a text parameter file. None reads it from a PTT archive and otherwise means "protein". Ignored when model is already a DCAModel. Must contain "-".
  • pseudocount: Mean-field only: pseudocount in [0, 1] that regularizes the covariance matrix before inversion.
  • device: "auto" (CUDA when available, else CPU), "cpu", "cuda" or "mps". Ignored for an in-memory model.
  • dtype: "float32" or "float64" for loaded parameters. Ignored for an in-memory model.

Returns: A ContactMapResult whose scores is a symmetric (L, L) array with a zero diagonal.

Raises:

  • InputValidationError: If both or neither of model and fasta_path are given, the alphabet lacks the "-" gap token, or the pseudocount is outside [0, 1].

Example: 

result = predict_contacts(model="params.dat.gz", alphabet="rna")
top = result.to_long_dataframe().nlargest(20, "score")  # 0-based positions

function compute_contact_map

compute_contact_map(
    model: 'DCAModel | str | Path | None' = None,
    fasta_path: 'AlignmentInput | None' = None,
    alphabet: 'str | None' = None,
    pseudocount: 'float' = 0.5,
    device: 'str' = 'auto',
    dtype: 'str' = 'float32'
) → ndarray

Return only the (L, L) contact-score matrix of predict_contacts.

Takes the same arguments as predict_contacts; see it for details.

Returns: A symmetric NumPy array of APC-corrected scores with a zero diagonal.

This file was automatically generated via lazydocs.