Integrating web data miningand machine learningalgorithms to predict progression free survival and overall survival in multiple myeloma patients

Abstract: Multiple myeloma patients have highly variable progression free survival and overall survivalranging from a few weeks to more than 5 years. Stratification of these patients can help toidentify high risk patients i-e patients with short-term progression free survival and overallsurvival. This background provides an opportunity for medical practitioners to treat patients atrisk with targeted therapies and improve their progression free survival and overall survival. Here, we have integrated a NetRank, a variation of the Google PageRank algorithm, withmachine learning algorithms to predict progression free survival and overall survival. For thispurpose, we have built a data set of newly diagnosed multiple myeloma patients (n=31)consisting of transcriptomic (features=28256), clinical (features=13), biochemical (features=12),and fluorescent in situ hybridization (features=3) data. The performance of prediction model wasevaluated using accuracy, precision, F1-score, and recall. We found that the final overall survival prediction model consisted of 29 features including 18transcriptomic, 3 clinical, 7 biochemical, and 1 fluorescent in situ hybridization features. Top 5 ofthese features were LDH, TMEM62, BUB1, PRPF18, and albumin while ISS and gender werethe least significant. Similarly, the final progression free survival model consisted of 21 featuresincluding 10 transcriptomic, 3 clinical, 7 biochemical, and 1 fluorescent in situ hybridizationfeatures. Top 5 of thesis features were TMEM62, Serum: M-protein, CDC42BPB, Hchain, andPLEKHM1 while SOX13, ISS, and 17p13 do not demonstrate significance. The findings are of paramount importance that prediction models built in this project may enablemedical practitioners to stratify the patients at risk as early as at the diagnosis stage and treatthem with targeted and personalized therapies to improve their overall survival and progressionfree survival.