Table 3.

Summary of benchmark datasets

Label	Species	Proteins	Interactions (positive/negative)	Prediction Tool
HS1	Homo sapiens	9439	37,027/37027	PRED_PPI (Guo et al.)
EC1	Escherichia coli	1834	6954/6954	PRED_PPI (Guo et al.)
DM1	Drosophila melanogaster	7059	21,975/21975	PRED_PPI (Guo et al.)
CE	Caenorhabditis elegans	2640	4030/4030	PRED_PPI (Guo et al.)
SC1	Saccharomyces cerevisiae	2245	3956/3956	PRED_PPI (Guo et al.)
HS2a	Homo sapiens	7033	24,718/177117	SPRINT (Li & Ilie)
HS3	Homo sapiens	1515	12,244/12244	TRI_tool (Perovic et al.)
SC2	Saccharomyces cerevisiae	3291	15,238/15238	go2ppi-RF (Maetschke et al.)
HS4	Homo sapiens	3296	3490/3490	go2ppi-RF (Maetschke et al.)
EC2	Escherichia coli	589	1167/1167	go2ppi-RF (Maetschke et al.)
SP	Schizosaccharomyces pombe	904	742/742	go2ppi-RF (Maetschke et al.)
AT	Arabidopsis thaliana	756	541/541	go2ppi-RF (Maetschke et al.)
MM	Mus musculus	1088	500/500	go2ppi-RF (Maetschke et al.)
DM2	Drosophila melanogaster	658	321/321	go2ppi-RF (Maetschke et al.)
SC3	Saccharomyces cerevisiae	2152	3844/3844	go2ppi-RF (Maetschke et al.)
HS5	Homo sapiens	6037	1091/3427	HVSM (Zhang et al.)
SC4	Saccharomyces cerevisiae	5436	4529/10831	HVSM (Zhang et al.)
SC5	Saccharomyces cerevisiae	454	500/500	GIS-MaxEnt (Armean et al.)
SC6	Saccharomyces cerevisiae	4424	17,257/48594	DeepSequencePPI (Gonzalez-Lopez et al.)

^aIn the work of SPRINT [6], the authors prepared three separate data into three human PPI data sets (i.e. C1, C2 and C3). To facilitate 10-fold CV in our experiments, we merged all three data sets into one single set of human PPI data with the redundancies removed