Development and validation of a comprehensive machine learning framework for a diagnostic model of uremia based on genes involved in major depressive disorder.

Background: Major depressive disorder (MDD) and uremia are two chronic wasting diseases that have interactive effects and significantly aggravate patients' distress. However, the molecular basis linking these diseases remains poorly investigated. Methods: Various machine learning algorithms were used to analyze transcriptome data from the Gene Expression Omnibus (GEO) datasets, including those from MDD and uremia patients, to develop and validate our model. After removing batch effects, differentially expressed genes (DEGs) were identified between each disease group and the control group. Functional enrichment analysis was then performed at the intersection of DEGs from the two diseases. In addition, single-sample gene set enrichment analysis (ssGSEA) quantitative immune infiltration analysis was conducted. The optimal diagnostic model of uremia was constructed by analyzing and verifying the training set with multiple combinations of 12 machine learning algorithms. Finally, potential drugs for uremia were identified using the "Enrichr" platform. Results: According to enrichment analysis, a total of seven key genes closely related to MDD and uremia, mainly involved in the immune process, were identified. Immune infiltration analysis showed that MDD and uremia had different profiles of immune cell infiltration compared to healthy controls. Powerful diagnostic markers of seven genes (IL7R, CD3D, RETN, RAB13, TNNT1, HP, and S100A12) were constructed from these genes, and all showed better performance than published uremia diagnostic models. In addition, decitabine and nine other agents were found to be potential agents for the treatment of uremia. Conclusion: Our study combined bioinformatics techniques and machine learning methods to develop a diagnostic model for uremia, focusing on common genes between MDD and uremia.