Identification and immune landscape of sarcopenia-related molecular clusters in inflammatory bowel disease by machine learning

Abstract Sarcopenia, a prevalent comorbidity of inflammatory bowel disease(IBD), is characterized by diminished skeletal muscle mass and strength, and it has garnered increasing attention. Nevertheless, the underlying interconnected mechanisms remain elusive. This study pioneered the identification of distinct expression patterns within sarcopenia-related genes (SRGs) across individuals with IBD and in samples of normal tissue. By analyzing various SRG expression profiles, we effectively segregated 860 IBD samples into two distinct clusters, each marked by its unique immune landscape. To unravel the transcriptional disruptions underlying these clusters, the Weighted Gene Co-expression Network Analysis (WGCNA) algorithm was employed to spotlight key genes linked to each cluster. Leveraging machine learning, namely RF, LASSO, and SVM-RFE algorithms, we established a diagnostic model grounded in 13 key genes (LYN, IFITM2, ACSL4, CLEC4E, SOCS3, PLAU, TIMP1, NCF2, MNDA, IL1B, CXCL1, MMP1, and S100A8). Moreover, the GSE112366 dataset facilitated the exploration of gene expression dynamics within the ileum mucosa of UC patients pre- and post-Ustekinumab treatment. Additionally, insights into the intricate relationship between immune cells and these pivotal genes were gleaned from the single-cell RNA (scRNA) dataset GSE162335. In conclusion, our findings collectively underscored the pivotal role of sarcopenia-related genes in the pathogenesis of IBD. Their potential as robust biomarkers for future diagnostic and therapeutic strategies is particularly promising, opening avenues for a deeper understanding and improved management of these interconnected conditions.