ADMET evaluation in drug discovery. 20. Prediction of breast cancer resistance protein inhibition through machine learning

Table 8 The representative fragments whose counts are equal or larger than 2 in the 50 misclassified compounds

Name	Count1^a	Count2^a	Count3^a	OR1^a (%)	OR2^a (%)	OR3^a (%)
Pyridine	125	43	5	5.58	7.69	10.00
Tetrahydro-2H-pyran	37	6	3	1.65	1.07	6.00
Adamantane	6	3	2	0.27	0.54	4.00
(E)-prop-1-ene-1,3-diyldibenzene	58	16	3	2.59	2.86	6.00
1H-indole	82	25	3	3.66	4.47	6.00
4H-chromene	110	17	4	4.91	3.04	8.00
Furan	175	37	2	7.81	6.62	4.00
Piperazine	98	22	3	4.38	3.94	6.00
Quinazoline	150	39	2	6.70	6.98	4.00
Quinoline	58	16	3	2.59	2.86	6.00
Thiophene	109	25	3	4.87	4.47	6.00
1,2,3,4-tetrahydroi	56	14	3	2.50	2.50	6.00
2-(4-((quinolin-3-ylmethyl)amino)phenethyl)-1,2,3,4-tetrahydroi	0	2	2	0.00	0.36	4.00

^aCount1, 2, and 3 represent the number of the training compounds, testing compounds and misclassified compounds containing the fragment, respectively, and OR1, 2 and 3 represent the occurrence ratio of the fragment in the training compounds, testing compounds and misclassified compounds, respectively

ISSN: 1758-2946