Single‐Cell and Bulk RNA‐Sequencing Unveils the Hub Biomarkers and Their Potential Mechanisms in Diabetic Erectile Dysfunction via Machine Learning

In this article, we firstly integrated scRNA‐sequencing and bulk RNA‐sequencing data to mine hub biomarkers and to clarify the microenvironment or mechanisms of diabetic erectile dysfunction (DMED) from the perspective of genomics. Bulk RNA‐sequencing and scRNA‐sequencing data for DMED were acquired from the GSE2457 and GSE206528 datasets, respectively. Hub biomarkers were identified by the machine learning method, and their potential mechanisms were also revealed in DMED. Based on these differently expressed genes in DMED, four different kinds of machine learning methods were simultaneously applied and compared. Therein, random forest (RF) was the optimal machine learning method, and its top five genes (Slco1a6, Angptl4, Mepe, Cidea, and Ubd) were regarded as hub biomarkers in DMED. These five hub biomarkers’ PPI networks, chemical‐gene‐ED networks, TF‐gene‐ED networks, RBP‐gene‐ED networks, and LncRNA/miRNA/mRNA networks, were also displayed by us for their potential mechanisms in DMED. By scRNA‐sequencing data analysis, our results showed that Angptl4 and Cidea were highly expressed in DMED compared to the control group. Moreover, Angptl4 was expressed in five annotated cell clusters, while Cidea was mainly expressed in the fibroblast (FB) cluster and the endothelial cell (EC) cluster. Further experimental results confirmed that Angptl4, Cidea, Mepe, and Slco1a6 were consistent in both the bulk RNA‐seq data analysis of GSE2457 and our qRT‐PCR results (all p < 0.05).In conclusions, Angptl4, Cidea, Mepe, and Slco1a6 could be hub biomarkers, providing novel targets for DMED in the future, and their involved mechanisms could provide novel insights into the DMED microenvironment.