Identification and experimental validation of autophagy-related genes in focal segmental glomerulosclerosis by integrating bioinformatics and machine learning

Focal segmental glomerulosclerosis (FSGS) is a common chronic glomerular disease characterized by podocyte injury. The aim of the present study was to investigate autophagy-related characteristics in FSGS. GSE200828, GSE99340, GSE47183, GSE108109, and GSE104948 were used as training sets, and GSE129973 was used as the validation set. 222 autophagy-related genes (ARGs) were incorporated. Autophagy-related differentially expressed genes (ARDEGs) were obtained and then analyzed using violin plots, PPI network analysis, functional enrichment analysis, gene set enrichment analysis, immune cell infiltration and cMAP. Two machine learning methods were applied to select candidate ARGs. Nomogram and receiving operating curve were conducted to assess diagnostic value and screen out core genes. An in vivo rat model of FSGS was established to verify expression of core genes via WB and qRT-PCR. The results indicated that nine ARDEGs were identified. PPI network contained nine nodes and 30 edges, and TP53 had the highest degree value. ARDEGs were significantly enriched in physiological processes and pathways related to autophagy and immunity. Patients with FSGS had higher levels of resting natural killer cells, monocytes and activated dendritic cells, and lower levels of plasma cells, follicular helper T cells, resting dendritic cells and resting mast cells. Through cMAP analysis, 10 small molecule compounds were identified which might work as potential therapeutic drugs in FSGS. Next, three candidate ARGs were obtained which were further evaluated by nomogram and diagnostic value. Among them, TP53 and RELA had high diagnostic values. In vivo, TP53 and RELA were at higher levels in FSGS than in the control group. In conclusion, TP53 and RELA are promising autophagy-related diagnostic and therapeutic markers in FSGS.