Hit compounds and count of their “active”/“inactive” docked poses as predicted based on GA/SHAP selected features (Table 4).

Hitsa	NCI Code	Captured Byb	Predicted number of active and inactive docked poses
			Xgboost			RF
			Active poses	Inactive poses	Percent active posesc	Active poses	Inactive poses	Percent active posesc
317	3590	1,2	183	31	85.5	78	30	72.2
318	20 261	1	40	77	34.2	ND	ND	ND
319	59 407	1	205	34	85.8	ND	ND	ND
320	65 832	1	65	56	53.7	ND	ND	ND
321	745 104	1,2	3	3	50.0	5	1	33.3
322	72 868	1	56	45	55.4	ND	ND	ND
323	77 028	1	52	64	44.8	ND	ND	ND
324	77 029	1	47	64	42.3	ND	ND	ND
325	82 523	1	152	20	88.4	ND	ND	ND
326	98 711	1	48	71	40.3	ND	ND	ND
327	100 791	1	70	38	64.8	ND	ND	ND
328	107 137	1	102	41	71.3	ND	ND	ND
329	107 139	2	ND	ND	ND	45	11	80.4
330	267 431	1,2	40	48	45.5	48	37	56.5
331	289 523	1	76	28	73.1	ND	ND	ND
332	338 310	2	ND	ND	ND	129	34	79.1
333	341 076	1,2	217	39	84.8	189	29	86.7
334	341 077	1,2	209	41	83.6	180	37	82.9
335	363 007	2	ND	ND	ND	119	42	73.9
336	372 667	1,2	158	26	85.9	131	40	76.6
337	373 233	1	53	55	49.1	ND	ND	ND
338	380 962	1	45	33	57.7	ND	ND	ND
339	645 793	2	ND	ND	ND	40	23	63.5
340	651 016	2	ND	ND	ND	52	33	61.2
341	669 269	1,2	135	54	71.4	124	33	79.0
342	722 969	2	ND	ND	ND	51	20	71.8

^aChemical structures are shown in Fig. S3.

^b1 represents XGboost-GFA model or Hypo-1, 2 represents RF-GFA model or Hypo-2.

^cDetermined by dividing the number of active poses by the total number of poses (active + inactive). ND: not determined.