Open-source platform to benchmark fingerprints for ligand-based virtual screening

Table 1 Pairwise post-hoc Friedman tests

	TT	AP	ECFC4	RDK5	Avalon	lAvalon	FCFP4	lECFP4	FCFC4	lECFP6	ECFP4	ECFP6	MACCS	ECFC0	Rank
TT		X	X	X	X	-	-	-	-	-	-	-	-	-	1
AP			X	X	X	X	o	o	-	-	-	-	-	-	1
ECFC4				X	X	X	X	X	o	o	o	-	-	-	1
RDK5					X	X	X	X	X	o	o	-	-	-	1
Avalon						X	X	X	X	o	o	-	-	-	1
lAvalon							X	X	X	X	X	X	-	-	1
FCFP4								X	X	X	X	X	-	-	1
lECFP4									X	X	X	X	-	-	1
FCFC4										X	X	X	o	-	1
lECFP6											X	X	o	o	1
ECFP4												X	o	o	1
ECFP6													X	o	1
MACCS														X	1
ECFC0															1
	lECFP4	TT	lECFP6	ECFP4	ECFC4	ECFP6	FCFP4	RDK5	AP	Avalon	lAvalon	FCFC4	MACCS	ECFC0	Rank
lECFP4		X	X	X	X	o	o	o	-	-	-	-	-	-	1
TT			X	X	X	X	X	X	o	o	-	-	-	-	1
lECFP6				X	X	X	X	X	o	o	o	-	-	-	1
ECFP4					X	X	X	X	X	X	X	-	-	-	1
ECFC4						X	X	X	X	X	X	-	-	-	1
ECFP6							X	X	X	X	X	o	-	-	1
FCFP4								X	X	X	X	-	-	-	1
RDK5									X	X	X	o	-	-	1
AP										X	X	o	-	-	1
Avalon											X	X	-	-	1
lAvalon												X	-	-	1
FCFC4													-	-	1
MACCS														X	13
ECFC0															13

Results from pairwise post-hoc Friedman tests of the average rank between 14 2D fingerprints for the evaluation methods AUC (top) and BEDROC(20) (bottom). Pairs of fingerprints with no statistically significant difference are marked with “X”, pairs with an adjusted p-value distribution around the confidence level α with “o”, and pairs with a statistically significant difference with “-”. Fingerprints are ordered according to ascending average rank.

ISSN: 1758-2946