In Silico Framework of Colon Cancer Classification based on Gene Expression with AI

ABSTRACT Colon cancer ranks as the third most significant contributor to cancer-related mortality on a global scale. Given that colon cancer is typically genetic disease, the utilization of microarray gene expression-based cancer classification holds significant importance within the realm of diagnostic approaches. Current challenges in cancer diagnosis include imbalanced data issues and biologically significant genes in microarray gene expression data, detecting colon cancer with better accuracy. This study aimed to handle imbalanced data and significant genes from thousands of genes to classify colon cancer with highest accuracy. Firstly, the SMOTE-ENN-LR method has been applied in this study to enhance the oversampling of minority class, and remove the noise from gene expression data for colon cancer classification. Secondly, LASSO-LR method also implemented on same dataset. In the testing phase, the proposed SMOTE-ENN-LR method has revealed an overall prediction accuracy of 96.67%, precision of 96.67%, recall of 97.50%, and F1-score of 98.70% in the testing stage for colon cancer. LASSO-LR method obtained 90.32% of accuracy. The proposed SMOTE-ENN-LR method outperformer than the other methodology investigated throughout the testing phase.