Abstract

Multiple myeloma (MM) is a clonal cell cancer characterized by excessive cell division of plasma cells in the bone marrow, which can then overcrowd healthy cells. As a result, end organ damage to kidneys, bones, and the liver occurs. The worldwide incidence of MM amounted to 160,000 cases in 2018 and 106,000 patients have succumbed to the disease. MM is diagnosed relatively well by detecting M monoclonal protein produced from cancerous cells, yet mortality rates remain high because there is a lack of a specific treatment. By identifying upregulated genes found in malignant plasma cells, scientists can develop new and stronger therapies tailored to potential driver genes. This study takes a novel machine learning approach to identify driver genes of MM. A single-cell RNA sequencing dataset obtained from Gene Expression Omnibus containing data from 29,367 plasma cells and 22,088 genes was utilized in this study. This study evaluated the performance of three machine learning models: Random Forest (RF), Support Vector Machine (SVM) and K-Nearest Neighbors (KNN), with RF achieving the highest accuracy of 95.61%.To name a few genes, the models identified ANKRD28 and HLA-DPA1 as potential driver genes that have been cross-validated with previous literature. Notably, the models identified RP5-1171I10.5–a gene not yet established to be associated with multiple myeloma which shows potential to be further studied for research. These genes show potential to be further studied for specific targeted genetic therapy.