Table 4.

Performance comparison. Here protein complexes are detected from three PPI datasets and CYC2008 is used as benchmark

Methods	Small			Large			Total
	Recall	Precision	F1	Recall	Precision	F1	Recall	Precision	F1
(a) Gavin et al.
MCODE	0.023	0.143	0.039	0.457	0.558	0.502	0.255	0.577	0.354
RNSC	0.234	0.208	0.221	0.567	0.453	0.504	0.453	0.356	0.399
DPClus	0.162	0.198	0.178	0.622	0.425	0.505	0.438	0.399	0.418
CORE	0.239	0.186	0.209	0.134	0.484	0.210	0.312	0.360	0.334
ClusterONE	0.072	0.517	0.127	0.567	0.580	0.574	0.347	0.707	0.465
CPredictor	0.095	0.365	0.150	0.575	0.517	0.545	0.384	0.624	0.475
CPredictor2.0	0.342	0.273	0.304	0.535	0.649	0.587	0.481	0.562	0.518
(b) Krogan et al.
MCODE	0.023	0.200	0.040	0.291	0.681	0.408	0.146	0.653	0.239
RNSC	0.491	0.163	0.244	0.465	0.550	0.504	0.599	0.272	0.374
DPClus	0.414	0.179	0.250	0.520	0.512	0.516	0.564	0.306	0.397
CORE	0.414	0.172	0.243	0.079	0.357	0.129	0.444	0.283	0.346
ClusterONE	0.176	0.186	0.180	0.528	0.181	0.269	0.499	0.308	0.381
CPredictor	0.243	0.282	0.261	0.488	0.549	0.517	0.447	0.516	0.479
CPredictor2.0	0.410	0.295	0.343	0.441	0.657	0.528	0.516	0.535	0.525
c) Collins et al.
MCODE	0.027	0.167	0.047	0.512	0.733	0.603	0.258	0.712	0.379
RNSC	0.401	0.315	0.353	0.575	0.707	0.634	0.556	0.499	0.526
DPClus	0.369	0.329	0.348	0.591	0.595	0.593	0.547	0.513	0.530
CORE	0.401	0.313	0.351	0.315	0.791	0.450	0.473	0.512	0.491
ClusterONE	0.320	0.392	0.352	0.614	0.549	0.580	0.550	0.587	0.568
CPredictor	0.257	0.449	0.327	0.598	0.652	0.624	0.473	0.657	0.550
CPredictor2.0	0.437	0.356	0.392	0.559	0.699	0.621	0.556	0.644	0.597

Each bold value means the largest performance measure among the compared methods on the given PPI dataset