Predictive Power of Machine Learning Models for Relapse Outcomes in Acute Myeloid Leukemia: Unveiling Key Genes and Pathways for Improved Patient Management

Abstract

Acute Myeloid Leukemia (AML) is a challenging form of blood cancer requiring accurate relapse prediction for effective therapy and patient management. In this study, we applied multiple machine learning techniques to a dataset of AML patients in order to develop a reliable model for predicting relapse and guiding treatment decisions. We utilized various feature selection methods to identify the most relevant features associated with relapse. Additionally, we investigated gene ontology using the Gene Ontology (GO) database to gain insights into the biological processes and KEGG pathways related to the selected features. Our findings revealed key genes and pathways implicated in AML relapse. Among the machine learning models, Decision Tree (DT) showed the highest accuracy in predicting relapse outcomes. Furthermore, we compared the performance of DT models across different feature selections, highlighting the significance of specific factors such as MCL1, WBC, HGB, and BAD.p112 in relapse prediction. The results of our study have important implications for tailoring treatment plans and improving patient outcomes in AML. By accurately identifying patients at high risk of relapse, our model can aid in early interventions and personalized therapies. Ultimately, our research contributes to advancing the field of machine learning in AML and lays the foundation for developing effective strategies to combat relapse in this disease.