BackgroundMultiple myeloma (MM) remains an essentially incurable disease. This study aimed to establish a predictive model for estimating prognosis in newly diagnosed MM based on gene expression profiles.MethodsRNA-seq data were downloaded from the Multiple Myeloma Research Foundation (MMRF) CoMMpass Study and the Genotype-Tissue Expression (GTEx) databases. Weighted gene coexpression network analysis (WGCNA) and protein-protein interaction network analysis were performed to identify hub genes. Enrichment analysis was also conducted. Patients were randomly split into training (70%) and validation (30%) datasets to build a prognostic scoring model based on the least absolute shrinkage and selection operator (LASSO). CIBERSORT was applied to estimate the proportion of 22 immune cells in the microenvironment. Drug sensitivity was analyzed using the OncoPredict algorithm.ResultsA total of 860 newly diagnosed MM samples and 444 normal counterparts were screened as the datasets. WGCNA was applied to analyze the RNA-seq data of 1589 intersecting genes between differentially expressed genes and prognostic genes. The blue module in the PPI networks was analyzed with Cytoscape, and 10 hub genes were identified using the MCODE plug-in. A three-gene (TTK, GINS1, and NCAPG) prognostic model was constructed. This risk model showed remarkable prognostic value. CIBERSORT assessment revealed the risk model to be correlated with activated memory CD4 T cells, M0 macrophages, M1 macrophages, eosinophils, activated dendritic cells, and activated mast cells. Furthermore, based on OncoPredict, high-risk MM patients were sensitive to eight drugs.ConclusionsWe identified and constructed a three-gene-based prognostic model, which may provide new and in-depth insights into the treatment of MM patients.