Molecular complex detection in protein interaction networks through reinforcement learning

BMC Bioinformatics

Table 1 RL algorithm has strong performance on a synthetic toy dataset

	FMM Precision	FMM Recall	FMM F-score	CMMF	UnSPA	Qi et al. F1 score	SPA	F-Grand K-Clique	F-weighted K-Clique
RL Algorithm	0.963	0.963	0.963	0.963	0.969	1.00	0.959	1.00	1.00
Super.Complex	0.999	0.999	0.999	1.00	0.999	1.00	0.998	1.00	1.00

The algorithm was trained on 7 toy complexes from a synthetic network of 62 nodes and 78 edges. It predicted 14 complexes which are evaluated against the 14 true complexes
FMM, F-similarity-based Maximal Matching; CMMF, Community-wise Maximum F-similarity-based F-score; UnSPA, Unbiased Sn-PPV Accuracy; SPA, Sn-PPV Accuracy

ISSN: 1471-2105