Table 1 DeepNull improves association results relative to the Baseline model on ten phenotypes from the UK Biobank.

From: DeepNull models non-linear covariate effects to improve phenotypic prediction and association power

Pheno

n

#Hits

%Improve

#Loci

%Improve

  

Baseline

DeepNull

 

Baseline

DeepNull

 

ALP

416,232

1697

1759

3.65%

336

350

4.17%

ALT

416,057

371

379

2.16%

173

174

0.58%

AST

414,743

337

351

4.15%

137

145

5.84%

ApoB

414,639

1172

1219

4.01%

200

217

8.50%

Calcium

381,934

726

739

1.79%

272

281

3.31%

GRP

65,896

28

38

35.71%

26

38

46.15%

LDL

415,892

950

993

4.53%

193

202

4.66%

Phosphate

381,362

658

664

0.91%

224

229

2.23%

SHBG

378,459

1084

1120

3.32%

319

323

1.25%

TG

416,295

1221

1254

2.70%

261

266

1.92%

Avg.

370,151

824.4

851.6

6.29%

214.1

222.5

7.86%

  1. n is the sample size, hits refers to the number of independent genome-wide significant associations detected, and loci is the number of independent regions after merging hits within 250 kb. Bold values in the table indicate the best results.
  2. Phenotypic abbreviations: ALP alkaline phosphatase, ALT alanine aminotransferase, AST aspartate aminotransferase, ApoB Apolipoprotein B, GRP glaucoma referral probability, LDL low-density lipoprotein, SHBG sex hormone-binding globulin, TG triglycerides.