Table 3.

Tissue- and Cell-Type-Specific Functional Predictions

Method	AUROC	AUPR
emVars in Tewhey et al.,76E116
FUN-LDA	0.707	0.468
GenoSkyline	0.673	0.394
ChromHMM	0.669	0.420
Segway	0.622	0.356
IDEAS	0.645	0.321
DNase	0.718	0.540
DNase-narrow	0.666	0.406
DNase-gapped	0.659	0.335
cepip_cell	0.653	0.321
cepip_combined	0.642	0.373
Regulatory Motifs in Kheradpour et al.,77E118 and HepG2
FUN-LDA	0.691	0.445
GenoSkyline	0.629	0.331
ChromHMM	0.606	0.344
Segway	0.618	0.334
IDEAS	0.546	0.290
DNase	0.719	0.506
DNase-narrow	0.561	0.312
DNase-gapped	0.550	0.291
cepip_cell	0.592	0.300
cepip_combined	0.641	0.364
Regulatory Motifs in Kheradpour et al.,77E123 and K562
FUN-LDA	0.645	0.287
GenoSkyline	0.620	0.256
ChromHMM	0.634	0.263
Segway	0.585	0.241
IDEAS	0.615	0.231
DNase	0.656	0.337
DNase-narrow	0.524	0.191
DNase-gapped	0.565	0.205
cepip_cell	0.606	0.217
cepip_combined	0.625	0.247
dsQTLs in Degner et al.,78E116
FUN-LDA	0.750	0.374
GenoSkyline	0.740	0.368
ChromHMM	0.639	0.303
Segway	0.580	0.277
IDEAS	0.677	0.330
DNase	0.823	0.474
DNase-narrow	0.665	0.345
DNase-gapped	0.662	0.313
cepip_cell	0.741	0.379
cepip_combined	0.760	0.398
deltaSVM	0.751	0.589
dsQTLs and eQTLs in Degner et al.,78E116
FUN-LDA	0.793	0.476
GenoSkyline	0.756	0.372
ChromHMM	0.721	0.403
Segway	0.648	0.340
IDEAS	0.700	0.334
DNase	0.832	0.529
DNase-narrow	0.713	0.376
DNase-gapped	0.701	0.327
cepip_cell	0.753	0.381
cepip_combined	0.769	0.473
deltaSVM	0.708	0.509

AUROC and AUPR values for discriminating between variants likely to be functional and control variants are shown. Results are shown for several datasets (three different cell lines) with experimental validation (MPRA) of potential regulatory variants and one dsQTL dataset (dsQTLs and eQTLs contains a subset of dsQTLs that are also eQTLs). Methods include FUN-LDA, GenoSkyline, ChromHMM (25-state model), Segway, IDEAS, DNase (quantitative, DNase-narrow, and DNase-gapped), cepip, and deltaSVM (note that deltaSVM predictions are available only for the dsQTL dataset).