An explainable predictive machine learning model reveals ARRB2 as a key gene in post-traumatic stress disorder: A GEO database study.

A number of processes and pathways have been reported in the development of Post-Traumatic Stress Disorder (PTSD), however, novel biomarkers need to be identified for a better diagnosis and management. We used the Limma package to identify differential genes, combined with weighted gene co-expression network analysis (WGCNA) and five machine learning algorithms to screen, and finally obtained five key genes. We used eight machine learning algorithms to construct the predictive model, and the results demonstrated that the support vector machine (SVM) algorithm had the best predictive efficiency, with an area under curve (AUC) value of 0.894303363. To clarify the decision-making process of the model, we used the Shapley Additive exPlanations (SHAP) method to rank the importance of the model's display features on all model samples. An immune-cell infiltration analysis revealed significant differences in the relative abundances of immune cells between controls and PTSD patients and a correlation with the key genes. Then we confirmed the expression of these five biomarkers in stress-related brain regions (prefrontal cortex (PFC), hippocampus (HIP), amygdala (AMY)) of Single prolonged stress and electric foot shock (SPS&S) rats through a series of experimental validation, and found that arrestin beta 2 (ARRB2) gene expression was significantly down-regulated in HIP, and verified the expression of ARRB2 in HIP by further Western Blotting as well as immunofluorescence experiments. In conclusion, machine learning and bioinformatics analysis along with experimental techniques identified ARRB2 as potential biomarkers for PTSD.