Machine Learning-Assisted Prognosis of Multiple Myeloma Side Population Cells via SRGs and OCLR Stemness Index

Multiple myeloma (MM) relapse is significantly influenced by cancer stem cells, particularly side population (SP) cells, though their clinical prognostic value requires further elucidation. This study developed and validated a novel SP cell-derived stemness-related gene signature score for MM prognostication and therapeutic guidance. Analysis of differ\u0002entially expressed genes from MM SP cells (GSE109651) using Weighted Gene Co-expression Network Analysis identified a key stemness-associated gene module (MEblue). The stemness\u0002related gene scores (SRGs), derived from six core KEGG pathways within MEblue, were calculated via single-sample gene set enrichment analysis across TCGA and GEO cohorts, showing significant correlation with the mRNAsi stemness index (r = 0.62, p < 1e-82). High expression of the hsa05222 pathway, a key SRGs component, strongly predicted poor prognosis (HR = 12.765, p < 0.0001). Elevated hsa05222 scores were also linked to altered drug response, including increased sensitivity to bortezomib and resistance to topotecan, while TIDE analysis suggested potentially reduced benefit from immunotherapy in high-SRGs patients. To improve prognostic accuracy, an optimized weighted score (OptW) for these pathways was developed using a non-linear programming algorithm. The OptW model outperformed both the average-weighted SRGs and the best single pathway in training (TCGA) and validation (GSE24080, GSE57317) cohorts. This work systematically established and validated an SP cell-associated stemness scoring framework, demonstrating its potential as an independent prognostic biomarker and offering a theoretical basis for future personalized, stem cell\u2013targeted therapies in MM.