THERAPY RESISTANCE MULTI-OMICS MACHINE LEARNING PREDICTION OF CHEMOTHERAPY RESISTANCE IN SOLID TUMORS WITH KIDNEY AND CARDIAC COMORBIDITIES

Abstract

A practical problem in chemotherapy resistance of solid tumor patients is kidney and cardiac comorbidities that limit the doses and increase the risk of toxicity. This paper introduces a multi-omics machine learning framework to predict chemotherapy resistance in the solid tumor patient population, who may have a renal and cardiovascular comorbid profile, and with the aim of therapy. The proposed method is a combination of genomic, transcriptomic, proteomic, metabolomic, clinical, renal-function and cardiac-function and treatment-response parameters, which enables identification of high-risk patterns of resistance before and during chemotherapy. The framework will include multi-omics feature selection methods and supervised-learning models that will be used to uncover complex tumor–host interactions that will be linked to drug metabolism, tumor adaptation, organ tolerance, and treatment failure. The model can be used to stratify the patients into chemotherapy sensitive and chemotherapy resistant groups and to take account of treatment limitations due to comorbidities. The explainable AI features also add to the clinical interpretability, as biomarkers, comorbidities indicators and variables associated to treatment predicting resistance were identified. The proposed framework may enable oncologists to make more informed decisions about more effective and less toxic chemotherapy treatments and to reduce unnecessary toxicity of chemotherapy, and improve precision oncology treatment and outcomes for patients with medically complex solid tumors.