Genetic prediction profile for adalimumab response in Slovenian Crohn's disease patients.

Abstract

Response to anti-TNF therapy is crucial for life expectancy and life quality in patients with severe Crohn's disease. We investigated if a previously reported gene expression profile predictive for infliximab response could be also applied to adalimumab response in an independent cohort. Forty-seven Slovene Crohn's disease patients indicated for adalimumab therapy were enrolled in the study. Inflamed and non-inflamed colon biopsy samples were obtained during routine colonoscopy prior to adalimumab treatment. Response to adalimumab was measured with IBDQ. Gene expression in inflamed and non-inflamed colon biopsy samples was measured with RT-qPCR. Genotypes were extracted from previously available genotype data. Statistical analysis was performed with SPSS software. The R package e1071 was used to train bootstrap aggregated support vector machines (SVM). SVM prediction model analysis was used to analyze pooled, non-inflamed, and inflamed colon tissue datasets using IBDQ response after 4, 12, 20 and 30 weeks of adalimumab treatment. The bagging approach was used in an endeavor to obtain 100 % accuracy using 10 × 100 or 100 × 100 iterations. Average adalimumab response prediction accuracy is 75.5 % for pooled samples, 90.5 % for inflamed samples, and 100 % for non-inflamed samples. Moreover, models trained on selected SNPs from analyzed genes had an average accuracy of 92.8 %, confirming the involvement of genetic regions mapping the reported genes. Finally, using combined gene expression and SNP data we observed 100 % adalimumab response prediction accuracy for pooled, inflamed, and non-inflamed datasets. Our study supports the reported genetic anti-TNF response profile and extends it for adalimumab prediction.