Third-party implementation in Python of L3 algorithm in "Network-based prediction of protein interactions" on Nature Communication 2019, doi: https://doi.org/10.1038/s41467-019-09177-y, its GitHub (implemented in C++): kpisti/L3

The ppiLPred.py implemented the main L3 algorithm, and the example.py provided an example of using the algorithm. A minimum eample of PPI dataset is defined in the example.py.

Let n(x) returns a set of neighbor nodes of x, candidatePPIs as a set of non-existing undirected edges of the PPI dataset, with the data structure [[A, B], ..., [Y,Z]] where [A, B] is a candidate PPI.

def L3(candidatePPIs):
    PPIScores = {}
    for [X, Y] in candidatePPIs:
        U = n(X), V = n(Y)
        uvPair = [all possible adjacent uv node pairs in U and V]
        score = 0
        for [u, v] in uvPair:
            score += 1/sqrt(|n(u)|*|n(v)|)
        PPIScores[[X, Y]] = score
    return sort(PPIScores)