Variable selection is crucial in statistical problems involving functional data, as it enhances prediction accuracy by filtering out irrelevant features. While selecting the most relevant variables is important, focusing solely on relevance can result in redundancy, negatively affecting model efficiency. The minimum Redundancy Maximum Relevance (mRMR) method addresses this by balancing relevance and redundancy using mutual information to evaluate variable relationships. Herein, we evaluate the mRMR method on a dataset of a single-cell RNA sequencing from Parkinson’s disease (PD) patients. We analyze and compare two different classification algorithms in terms of their performance in predicting the target variable and the computational time required. Additionally, we investigate the performance of these algorithms on the same dataset without feature selection, analyzing and comparing the results.