Table 4.

Maximum F-measure ($F_{\mathit{max}}$) scores for precision and recall (PR) curves obtained with DomFun using the Pathway Prediction Performance benchmark procedure

Domains classification	FA	Association + combination methods
		HyI+Fis	PCC+Sto	Jac+Sto	Sim+Sto
$F_{\mathit{max}}$ values from PR curves (domains-FunFys associations from UniProt proteins)
FunFams	GOMF	0.779	0.749	0.749	0.850
	GOBP	0.643	0.604	0.604	0.714
	GOCC	0.750	0.704	0.704	0.824
	KEGG	0.730	0.730	0.730	0.822
	Reactome	0.762	0.680	0.663	0.822
Superfamilies	GOMF	0.241	0.370	0.373	0.139
	GOBP	0.196	0.291	0.305	0.089
	GOCC	0.266	0.221	0.217	0.129
	KEGG	0.132	0.340	0.271	0.494
	Reactome	0.127	0.344	0.327	0.081

 The best performing methods for each domain/annotation source  combination are indicated in bold

FA Functional annotation, HyI hypergeometric index, PCC Pearson correlation coefficient, Jac Jaccard index, Sim Simpson index, Sto Stouffer’s method, Fis Fisher’s method