Characterization of genomic patterns that predict long-term metabolic side effects of antipsychotic treatments

Abstract

Aripiprazole and olanzapine are atypical antipsychotics used mainly for schizophrenia treatment. Being schizophrenia a chronic disorder, it requires lifelong medication that subsequently triggers metabolic Adverse Drug Reactions (ADRs). Therefore, the project's main objective was to contribute to evaluating short- and long-term antipsychotic drug responses and their impact on the development of metabolic diseases. An open, controlled, randomized, crossover clinical trial in healthy volunteers was carried out to evaluate the short-term effect (5 days) of olanzapine and aripiprazole. Blood samples were collected before and after each treatment, making 48 samples. Mice models were treated with short (5-days) and long-term (6 months) aripiprazole and olanzapine schemes. Liver and pancreatic islets were isolated. Total RNA was extracted from the blood and tissue samples and processed with Illumina TruSeq Stranded RNASeq technology. Differential expression analyses were performed, and differentially expressed genes were queried against KEGG and Gene Ontology databases to perform Gene Set Enrichment Analyses. A machine learning approach was used to train a model to classify the volunteers’ samples according to their response to the treatments and, therefore, to predict the appearance of metabolic side effects. Feature selection strategies such as F-score and Random Forest were used to select the most relevant features for the classification of the samples. The analysis of the different models allowed us to propose that olanzapine-induced metabolic disarrangements may be explained by PEPCK signalling repression from low-grade inflammation that may occur due to TLR4 signalling activation. Although aripiprazole may induce inflammatory signalling, it seems to activate compensation mechanisms such as promoting fatty acid synthesis to avoid metabolic disarrangements. Differential response of the human volunteers to the treatments was identified, which may be at least partially attributed to polymorphisms in PPARGC1, which prevented the volunteers from activating aripiprazole’s compensatory phenotypes. Stratification of the volunteers according to their response was possible, which may allow the development of a predictive pharmacogenomic test for the patients’ response shortly after the beginning of the treatment, a step forward to personalized medicine