Genomics of human NAFLD: Lack of data reproducibility and high interpatient variability in drug target expression as major causes of drug failures

Abstract

Background and aims: NAFLD is a major disease burden and a foremost cause of chronic liver disease. Presently, nearly 300 trials evaluate the therapeutic efficacy of > 20 drugs. Remarkably, the majority of drugs fail. To better comprehend drug failures, we investigated the reproducibility of fatty liver genomic data across 418 liver biopsies and evaluated the interpatient variability of 18 drug targets.

Approach and results: Apart from our own data, we retrieved NAFLD biopsy genomic data sets from public repositories and considered patient demographics. We divided the data into test and validation sets, assessed the reproducibility of differentially expressed genes and performed gene enrichment analysis. Patients were stratified by disease activity score, fibrosis grades and sex, and we investigated the regulation of 18 drug targets across 418 NAFLD biopsies of which 278 are NASH cases. We observed poor reproducibility of differentially expressed genes across 9 independent studies. On average, only 4% of differentially expressed genes are commonly regulated based on identical sex and 2% based on identical NAS disease score and fibrosis grade. Furthermore, we observed sex-specific gene regulations, and for females, we noticed induced expression of genes coding for inflammatory response, Ag presentation, and processing. Conversely, extracellular matrix receptor interactions are upregulated in males, and the data agree with clinical findings. Strikingly, and with the exception of stearoyl-CoA desaturase, most drug targets are not regulated in > 80% of patients.

Conclusions: Lack of data reproducibility, high interpatient variability, and the absence of disease-dependent drug target regulations are likely causes of NASH drug failures in clinical trials.