Skip to main content
. 2019 Apr 29;78(8):1055–1061. doi: 10.1136/annrheumdis-2018-214877

Table 2.

Prediction of response to TNFi using penalised regional genotypic scores for different types of intermediate traits

Intermediate trait type No of regional scores No of filtered scores Prediction of
∆SJC (%)
Prediction of
∆ESR (%)
Rheumatoid arthritis 37 37 5.3 (0.26) −1.6 (0)
Immune cell traits 508 470 −0.7 (0) 2.9 (0.17)
eQTLs 94 87 3.4 (0.16) 2.9 (0.17)
eQTLs and mQTLs 268 228 2.5 (0.11) 1.6 (0.09)

Prediction performance is quantified by the difference in test log-likelihood (in nats) between a model with clinical covariates and genotypic scores and a model with clinical covariates only and by the per cent of phenotypic variance explained (in parenthesis). Results from 10-fold cross-validation.

eQTL, expression quantitative trait loci; mQTL, methylation quantitative trait loci.ESR, erythrocyte sedimentation rate; SJC, swollen joint count; TNFi, tumour necrosis factor inhibitors;