Cytochrome P450 site of metabolism prediction from 2D topological fingerprints using GPU accelerated probabilistic classifiers

Table 5 Benchmarking analysis

Data	Method	q	Runtime seconds			x-fold to Ref
set			Ref	GT640	C2075	GT640	C2075
	RASCAL	40	42	11	2.2	4	19
	NB	40	8.8	1.4	0.3	6	29
2C9	PRW	40	436	114	32	4	14
	NB	175	8.2	1.3	0.3	6	27
	PRW	175	7.0	31	3.6	0.2	2
	RASCAL	40	57	15	3.0	4	19
	NB	40	12	1.9	0.4	6	30
2D6	PRW	40	727	157	44	5	17
	NB	175	11	1.8	0.4	6	28
	PRW	175	9.7	43	5.0	0.2	2
	RASCAL	40	298	66	13	5	23
	NB	40	55	8.3	1.8	7	31
3A4	PRW	40	3,654	714	201	5	18
	NB	175	53	8.1	1.8	7	29
	PRW	175	45	197	23	0.2	2

Table shows the time taken in seconds and relative performance o the reference (Ref) to run all three classifiers over the three data sets using a scenario with bond depth of 6 (175 features) and j = 201 and q = 40. The reference is a single Intel Xeon E5506 at 2.13GHz. The number of vectors in data sets CYP 2C9, 2D6 and 3A4 are 2905, 3399 and 7294, respectively.

ISSN: 1758-2946