Precision Medicine in Psychiatry by Predicting Lithium Treatment Response using Gene Expression Biomarkers and Machine Learning

Abstract

BackgroundWe tested the hypothesis that transcriptional‐level genes signatures are differentially expressed between male and female bipolar patients, prior to lithium treatment, in a patient cohort who were later were clinically classified as lithium treatment responders using the primary outcome measure of the Clinical Global Impression Scale for Bipolar Disorder‐Severity (CGI‐BP‐S).MethodsGene expression study data was obtained from the Lithium Treatment‐Moderate dose Use Study data accessed from the National Center for Biotechnology Information's Gene Expression Omnibus via accession number GSE4548. Differential gene expression analysis using the CGI‐BP‐S responder status and Gender served as control variables for analysis using the Linear Models for Microarray and RNA‐Seq (limma) package and the Random Forests machine learning algorithm in R for Statistical Computing programming language software.ResultsIn pre‐treatment lithium responders, the following genes were found having a greater than 0.5 fold‐change and differentially expressed indicating a male bias: RBPMS2, SIDT2, CDH23, LILRA5, and KIR2DS5 while the female‐biased genes were identified as: HLA‐H, RPS23, FHL3, RPL10A, NBPF14, PSTPIP2, FAM117B, CHST7, and ABRACL.ConclusionsApplying machine learning methods to gene expression data, we developed a pre‐treatment gender‐specific gene‐expression‐based predictive model selective for lithium treatment responders according to the Clinical Global Impression Scale for Bipolar Disorder‐Severity with an ROC AUC of 0.92 for men and an ROC AUC of 1 for women.Support or Funding InformationNone.