Implementation of multiple-instance learning in drug activity prediction

BMC Bioinformatics

Table 3 Conformational sampling and pharmacophore fingerprints

Data set	n _training a	n _test b	m _{pre-filtering} c	m _{post-filtering} d	m _significant e	Δ^*f
I	17249	5399	1872521	243721	2979	1.77
II	35434	12333	1670985	155220	14002	5.40
III	32528	9942	1636254	145996	1542	1.80
IV	41960	10746	13687602	161018	3467	1.66

^a The number of conformers in the training set; ^b the number of conformers in the test set; ^c the number of pharmacophore bits in the fingerprint originally enumerated; ^d the number of pharmacophore bits in the fingerprint after filtering; ^e the number of bits in the optimal subset of the pharmacophore fingerprint; ^f the optimal threshold value to select truly significant pharmacophore bits.

ISSN: 1471-2105