Automating exosome-based glioblastoma diagnosis through bioinformatics and machine learning
This project aims to integrate bioinformatics tools together with machine learning approaches for the standardization of exosome-based diagnostic approaches for glioblastoma diagnosis. Bioinformatics tools show that differentially expressed miRNAs between patients with different glioma types and healthy controls are indeed involved in tumor progression-related processes. However, the variation in these differentially expressed miRNAs between different datasets deteriorate the diagnostic role of miRNA sets.
By using differentially expressed miRNAs between glioblastoma patients and healthy conrols as a feature set for a binary SVM model, the algorithm perfectly distinguishes glioblastoma patients from healthy control in both datasets, suggesting the superiority of exploiting whole miRNA profiling instead of some set of miRNAs for diagnostic approaches.