Screening of molecular signatures from microRNA and mRNA expression profiling related to the efficacy of the antidepressant duloxetine and analysis of their regulatory relationship

Previous clinical studies have reported that not all depressed patients respond to antidepressants. Therefore, finding potential predictive molecular biomarkers is crucial for providing important guidelines in the diagnosis and treatment of depression. In recent years, some studies have identified potential signature molecules using machine learning analysis of transcriptomic data. In this study, we analyzed transcriptomic data from duloxetine-treated depressed patients and used a random forest algorithm to identify a set of characteristic microRNAs and mRNAs. For the mRNA expression dataset, we identified a gene set having 30 transcripts, including TNK2 (Gini index =2.11) and KDM2A (Gini index =1.81), with an AUC value 10-fold and cross-validation equal to 0.907. We also identified the feature microRNA sets (n = 20) with an AUC value of 0.711. Gene annotation function analysis suggests that those feature mRNAs mainly mediate myosin complex and EGFR-related activities. When we explored the relationship between microRNAs and mRNAs, 151 microRNA-mRNA pairs showed negative correlations, while 129 pairs showed positive correlations. Meanwhile, 58 microRNA-mRNA pairs showed potential functional regulatory relationships based on seed-region sequence complementarity. We also confirmed direct regulation between several characteristic microRNAs and mRNAs using luciferase reporter assays. Subsequently, we found that the expression levels of TNK2 (P = 0.0044) and KDM2A (P = 0.0080) were significantly up-regulated after duloxetine treatment of SYSH cells. In a depressive animal model, mice administered with duloxetine showed altered mRNA and protein expression of those feature genes in the prefrontal cortex. Altogether, we identified the signature microRNAs and mRNAs that can distinguish duloxetine-effective and duloxetine-ineffective phenotypes and a close regulatory relationship between signature microRNAs and mRNAs. These characteristic molecular signatures may serve as predictive biomarkers for duloxetine efficacy, enabling patient stratification before and during treatment and providing a theoretical foundation for individualized antidepressant therapy and informed clinical decision-making. This graphical abstract summarizes an integrative workflow combining transcriptomic profiling, machine learning–based feature selection, and experimental validation to identify duloxetine-related microRNA-mRNA regulatory signatures associated with antidepressant response. • Machine learning identified duloxetine-related microRNAs and mRNAs (e.g., TNK2 , KDM2A , miR-361-5p). • MicroRNA–mRNA regulatory interactions were validated by in silico analysis and luciferase assays. • Duloxetine reshaped the pharmacological network based on key microRNAs and mRNAs. • Signature microRNAs and mRNAs distinguish duloxetine responders from non-responders.