Evaluation of Machine Learning Algorithms for the Discovery of Tumor Treatments

Abstract

Machine learning is regularly used to interpret and analyze information from large and complex datasets originating from numerous fields. In one of those fields, namely Bioinformatics, the exploration of potentially beneficial drug configurations for tumor treatments via simulations requires multiple processing units to be used in parallel and a considerable amount of time to be completed. In this thesis, we apply a state-of-the-art model exploration active learning workflow for the characterization of a new drug configuration parameter space, using a redesigned simulator for in silico experiments. Moreover, we incorporate different clustering and optimization approaches in the original workflow and compare their performance in simulation trials on high-performance computing infrastructure. The overall goal is to divide the parameter space into regions that contain effective and ineffective treatments, and thus guide the related research towards more focused and effective real-world trials. Experimental results demonstrate that the workflow achieves a fine characterization of the treatment parameter space. Moreover, results indicate that the incorporation of different clustering and optimization algorithms in the workflow affects the quality of the treatment space characterization and the number of required simulations.