haddock.modules.scoring.deeprank.deeprank module

Wrapper around deeprank-gnn-esm for use as a HADDOCK3 scoring backend.

class haddock.modules.scoring.deeprank.deeprank.DeeprankWrapper(models, ncores, chain_i, chain_j)

Bases: object

Run deeprank-gnn-esm on a set of models and collect predicted scores.

ENSEMBLE_NAME = 'deeprank_ensemble.pdb'

run() → dict[str, float]

Score all models and return the predicted scores per model.

deeprank-gnn-esm writes its ensemble/graph/embedding intermediates next to the input file with no way to redirect them, and none of that is useful once the scores are parsed out. So the whole run (ensemble build, scoring, csv parsing) happens inside a temporary directory that is discarded on exit, and only the scores dict is returned to the caller.

haddock.modules.scoring.deeprank.deeprank.deeprank_is_available() → bool

Check whether deeprank-gnn-esm’s requirements are met.