Feature Selection for Gene Expression in COVID-19 Using Binary Slime Mold Algorithm and K-Nearest Neighbors

Identifying critical genes associated with COVID-19 is vital to advance our understanding and treatment of the disease. This study leverages the GSE149273 dataset, which encompasses gene expression profiles from individuals with respiratory viral infections, to enhance feature selection methodologies. Initially, the dataset contained 25343 genes. We introduce a novel approach employing the binary slime mold algorithm (BSMA) combined with k-nearest neighbors (KNN) for gene selection and classification. Our method achieved a remarkable classification accuracy of 95.56 % while utilizing an average of only 3.7 genes. The robustness and efficacy of our approach were validated by comparing its performance against six existing binary metaheuristic algorithms, where it consistently outperformed them in terms of accuracy and gene selection efficiency. The significant reduction in the number of genes required for accurate classification underscores the potential of our approach for simplifying diagnostic processes and reducing computational complexity.