Virtual and in vitro screening of approved drugs targeting genes involved in colorectal cancer: an approach with drug repurposing

Background and Aims: Colorectal cancer is the second leading cause of cancer death and the third most common cancer worldwide. In this research, given the lengthy process of discovering new drugs, machine learning was employed to select the compounds to inhibit specific signaling pathways associated with colorectal cancer, with the goal of repurposing FDA-approved drugs. Methods: The target gene was identified via the GEPIA database and literature review. As a metalloprotease target, it aids extracellular matrix degradation, promoting colorectal cancer progression and metastasis. Based on the IC۵۰ values of the tested compounds for the target genes in the ChEMBL database, two groups—active and inactive—were classified. The morgan fingerprint and physicochemical properties methods from the RDKit package was used to calculate the molecular descriptors. The data was split into the two groups: train and test with a ratio of ۸۰ to ۲۰. Using the training data, a machine learning model was trained to predict the effect of the FDA-approved drugs on the target gene. The machine learning model was developed using eight algorithms (k-nearest neighbors, Decision tree, Support vector machine, Naïve Bayes, Logistic regression, Random Forest, Multilayer Perceptron, Extreme Gradient Boosting). The models were tested, and accuracy, AUC, and other parameters were calculated. A total of ۲,۷۶۶ FDA-approved drugs were fed into the top models, and the results from the models' predictions were analyzed. Results: XGBOOST and SVM models were selected due to their highest accuracy and AUC. The accuracy of both models was ۹۲%, while the AUC values were ۹۸% and ۹۶%, respectively. In the prediction results of FDA approved drugs, drugs that were above ۹۵% active were selected and the two models' results were compared. Finally, we identified ۵۸ drugs out of ۲,۷۶۶. Some of the final drugs include: Bromocriptine , Ledipasvir , Telaprevir. Conclusion: Docking and molecular dynamics