Table 1. Classification performance for feature groups.

Performance characteristics of the full model are shown at the top, and analyses for individual groups of features are shown in sections below. Counts are shown for true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN). *Mean distance aggregates across all the variants in the credible set and weighs by their posterior probability.

Features	Average precision	AUC	Precision	Recall	TP	FP	TN	FN	Sensitivity	Specificity	FDR	GSP count	GSN count
Full model	0.65	0.93	0.73	0.53	236	86	6,429	209	0.53	0.99	0.27	445	6,515
Naïve closest gene classification
Closest footprint	0.37	0.79	0.56	0.60	268	207	6,308	177	0.60	0.97	0.44	445	6,515
Closest TSS	0.34	0.76	0.56	0.55	246	195	6,320	199	0.55	0.97	0.44	445	6,515
Leave-one-group-in
Mean distance*	0.62	0.91	0.69	0.49	219	98	6,417	226	0.49	0.98	0.31	445	6,515
Interaction	0.26	0.79	0.55	0.05	23	19	6,496	422	0.05	1.00	0.45	445	6,515
Molecular QTL	0.36	0.85	0.62	0.18	79	49	6,466	366	0.18	0.99	0.38	445	6,515
Pathogenicity prediction	0.48	0.76	0.70	0.43	191	80	6,435	254	0.43	0.99	0.30	445	6,515
Leave-one-group-out
Mean distance*	0.47	0.77	0.69	0.43	191	84	6,431	254	0.43	0.99	0.31	445	6,515
Interaction	0.65	0.93	0.73	0.53	234	85	6,430	211	0.53	0.99	0.27	445	6,515
Molecular QTL	0.65	0.93	0.74	0.54	239	86	6,429	206	0.54	0.99	0.26	445	6,515
Pathogenicity prediction	0.63	0.92	0.71	0.50	222	91	6,424	223	0.50	0.99	0.29	445	6,515