Skip to content

Recovery

recover_false_negatives(candidates: list[tuple[str, str]], dist_matrix: np.ndarray, node_index: dict[str, int], motif_scores: dict[tuple[str, str], float]) -> list[dict]

Recover kinase–substrate pairs missed by the motif step.

Parameters:

Name Type Description Default
candidates list[tuple[str, str]]
required
dist_matrix ndarray
required
node_index dict[str, int]
required
motif_scores dict of {(kinase_id, substrate_id): motif_score} from pynetphorest
required

Returns:

Type Description
List of prediction dicts for recovered pairs, each with:

kinase_id, substrate_uniprot, motif_score (-1.0 sentinel), context_score, networkin_score, recovered (True), recovery_method
Source code in src/pynetworkin/recovery.py 
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
def recover_false_negatives(
    candidates: list[tuple[str, str]],
    dist_matrix: np.ndarray,
    node_index: dict[str, int],
    motif_scores: dict[tuple[str, str], float],
) -> list[dict]:
    """
    Recover kinase–substrate pairs missed by the motif step.

    Parameters
    ----------
    candidates   : all (kinase_id, substrate_uniprot) pairs from the STRING network
    dist_matrix  : all-pairs shortest path matrix from floyd_warshall()
    node_index   : maps protein ID -> row/col index in dist_matrix
    motif_scores : dict of {(kinase_id, substrate_id): motif_score} from pynetphorest

    Returns
    -------
    List of prediction dicts for recovered pairs, each with:
        kinase_id, substrate_uniprot, motif_score (-1.0 sentinel),
        context_score, networkin_score, recovered (True), recovery_method
    """
    recovered = []
    for kinase_id, substrate_id in candidates:
        # Skip if motif step already scored this pair
        if (kinase_id, substrate_id) in motif_scores:
            continue

        ki = node_index.get(kinase_id)
        si = node_index.get(substrate_id)
        if ki is None or si is None:
            continue

        d = dist_matrix[ki, si]
        if d == np.inf:
            continue

        c_score = float(1.0 / (1.0 + d))
        if c_score >= CONTEXT_RECOVERY_THRESHOLD:
            recovered.append(
                {
                    "kinase_id": kinase_id,
                    "substrate_uniprot": substrate_id,
                    "motif_score": -1.0,  # sentinel: motif step did not score this pair
                    "context_score": c_score,
                    "networkin_score": c_score,  # context-only score for recovered pairs
                    "recovered": True,
                    "recovery_method": "context_proximity",
                }
            )

    return recovered