AnnoPRO: a strategy for protein function annotation based on multi-scale protein representation and a hybrid deep learning of dual-path encoding

Table 2 A comparison among those performances of AnnoPRO and two state-of-the-art methods (DeepGOPlus and PFmulDL) on predicting two groups of ‘Independent Testing’ data (SameSP and DiffSP)

	Method	BP		CC		MF
	Method	F_max	AUPRC	F_max	AUPRC	F_max	AUPRC
SameSP	DeepGOPlus	0.612	0.593	0.539	0.470	0.668	0.698
	PFmulDL	0.347	0.286	0.573	0.603	0.436	0.402
	AnnoPRO	0.610	0.589	0.759	0.772	0.835	0.829
DiffSP	DeepGOPlus	0.538	0.469	0.684	0.622	0.517	0.429
	PFmulDL	0.261	0.176	0.593	0.580	0.354	0.273
	AnnoPRO	0.602	0.552	0.742	0.741	0.749	0.739

SameSP had 1,859 proteins from 17 species covered by ‘Training’ and ‘Validation’ datasets of this study; DiffSP included 3,764 proteins from the remaining 997 species unique in ‘Independent Testing’ data of this study. Those values indicating the best performance among all three methods were highlighted in BOLD, and AnnoPRO performed the best in the vast majority of the Gene Ontology (GO) classes (BP, CC, MF) under both evaluating criteria (F_max, AUPRC). BP biological process, CC cellular component, MF molecular function

ISSN: 1474-760X