Identification of diagnostic molecules and potential therapeutic agents for atopic dermatitis by single-cell RNA sequencing combined with a systematic computing framework that integrates network pharmacology.

Atopic dermatitis (AD) is composed of highly flexible cellular participants. To better understand its pathobiology and molecular regulation mechanisms, it is necessary to combine single-cell RNA sequencing (scRNA-seq) with new computing frameworks or specific technologies, which may contribute to the development of better treatments for AD. The scRNA-seq data of GSE180885 and bulk RNA-seq data of GSE193309 were obtained from Gene Expression Omnibus (GEO) database, and the scRNA-seq data was analyzed by Seurat package to identify the cell types in AD. The genes related to the activity of AD topical drugs were obtained from the ChEMBL database, which provided a variety of bioactivity data such as multiple drugs and targets. AD-related genes were obtained from DisGeNET and CTD databases synthesizing human disease-related genes; the intersection of AD-related genes from these three sources with differentially expressed genes (DEGs) between non-diseased AD and normal human skin (NHS) samples and differential cell type marker genes was taken. The proximity analysis of drug gene network was performed based on the gene with the largest area of receiver operating characteristic (ROC) curve. Ten distinct cell types of AD and NHS were identified, except for phagocytes cells. Three hub genes, F10 and CALCRL and CTSB, were obtained. The area under the curve of ROC based on CTSB expression was the largest, which was 60.15%. By binding drug CTSB-related gene interaction network, we identified 145 potential drugs. Among them, the score of DB07045 and CTSB docking was the lowest, and molecular docking and molecular dynamics (MD) simulation confirmed the close and stable binding of DB07045 and cathepsin B. This work identified diagnostic molecules and potential therapeutic drugs of AD by scRNA-seq combined with a systematic computing framework of network pharmacology, which may provide valuable clues for drug design.