Abstract
Glioblastoma (GBM) is a fatal and aggressive form of brain cancer, described by rapid progression, poor prognosis, and limited treatment options. This study aims to apply a hybrid of two popular feature selection methods for the categorization of drug sensitivity features in GBM versus other cancer cell lines by employing a rank-based weighting combination scheme to identify discriminative drug compound feature sets. This approach is necessary to reduce dimensionality and enhance classification performance while increasing the interpretability of the prediction model. The experimental results indicate that the utilized machine learning (ML)-driven feature selection approach achieves more than 95% accuracy value and obtains less than or equal to 11 selected features for each drug sensitivity metric on Genomics of Drug Sensitivity in Cancer (GDSC) datasets with high-dimensional space. Our drug compound-based findings demonstrate that our feature selection approach improves model stability and performance, paving the way for more precise and clinically actionable advancements in GBM research.