A Framework Based on Metabolic Networks and Biomedical Images Data to Discriminate Glioma Grades

Collecting and integrating information from dierent datasources is a successful approach to investigate complex biological phe-nomena and to address tasks such as disease subtyping, biomarker pre-diction, target, and mechanisms identication. Here, we describe an in-tegrative framework, based on the combination of transcriptomics data,metabolic networks, and magnetic resonance images, to classify dierentgrades of glioma, one of the most common types of primary brain tu-mors arising from glial cells. The framework is composed of three mainblocks for feature sorting, choosing the best number of sorted features,and classication model building. We investigate dierent methods foreach of the blocks, highlighting those that lead to the best results. Ourapproach demonstrates how the integration of molecular and imagingdata achieves better classication performance than using the individualdata-sets, also comparing results with state-of-the-art competitors. Theproposed framework can be considered as a starting point for a clinicallyrelevant grading system, and the related software made available lays thefoundations for future comparisons.