AI Machine Learning Technique Characterizes Potential Markers of Depression in Two Animal Models of Depression.

Abstract

(1) Background: there is an urgent clinical need for rapid and effective antidepressants. (2) Methods: We employed proteomics to profile proteins in two animal models (n = 48) of Chronic Unpredictable Stress and Chronic Social Defeat Stress. Additionally, partial least squares projection to latent structure discriminant analysis and machine learning were used to distinguish the models and the healthy control, extract and select protein features and build biomarker panels for the identification of different mouse models of depression. (3) Results: The two depression models were significantly different from the healthy control, and there were common changes in proteins in the depression-related brain regions of the two models; i.e., SRCN1 was down-regulated in the dorsal raphe nucleus in both models of depression. Additionally, SYIM was up-regulated in the medial prefrontal cortex in the two depression models. Bioinformatics analysis suggested that perturbed proteins are involved in energy metabolism, nerve projection, etc. Further examination confirmed that the trends of feature proteins were consistent with mRNA expression levels. (4) Conclusions: To the best of our knowledge, this is the first study to probe new targets of depression in multiple brain regions of two typical models of depression, which could be targets worthy of study.