Similarity maps - a visualization strategy for molecular fingerprints and machine-learning methods

Table 1 Dice similarities and maximum weights

FP	s_Dice		Max. Weight		Method	PP		Max. Weight
	2	3	2	3		2	3	2	3
AP	0.604	0.531	0.028	0.033	RF(Morgan2)	0.950	0.600	0.660	0.370
Morgan2	0.561	0.381	0.123	0.110	NB(Morgan2)	1.000	0.999	12.5	20.92
CountMorgan2	0.599	0.529	0.091	0.049
FeatMorgan2	0.554	0.469	0.176	0.171

Dice similarity s_Dice and maximum weight between reference compound 1 and molecules 2 and 3 using the basic fingerprints (FP) atom pairs (AP), Morgan fingerprint as bit vector with radius 2 (Morgan2), Morgan fingerprint as count vector with radius 2 (CountMorgan2) and Feature Morgan fingerprint with radius 2 (FeatMorgan2). Predicted probability to be active (PP) and maximum weight for random forest (RF) and naïve Bayes (NB) for molecules 2 and 3 with basic fingerprint Morgan2. The bit vectors of the circular fingerprints had the size 1024 bits. The default maximum path length of 30 was used for AP.

ISSN: 1758-2946