Developing Web Application for Virtual Screening and Prediction of Breast Cancer Drugs: Target Interaction and Drug Approval Using Machine Learning

Breast cancer is the most prevalent cancer globally with a high patient mortality rate. As a result, researchers are trying to find new effective drugs to treat the disease. However, drug discovery is costly and time-consuming, making it difficult to succeed. Virtual screening is a tool for drug discovery as it can quickly assess the potential of many drug candidates, reducing time and cost. The study aims to develop a web application for the virtual screening of breast cancer drugs using a machine-learning approach. The data for this study was collected from the ChEMBL database and contained 8 human breast cancer target proteins, including mTOR, Trop-2, PI3K, CDK4/6, HER2, Aromatase, Akt, and Estrogen Receptor. The data is preprocessed with Feature Engineering using Lipinski’s rule with the SMILE structure of a compound. The data is also applied in developing machine learning algorithms with two main objectives: predicting Drug–Target interaction and Drug Approvals with a classification model. Results of the study show that the SMILE structure and Lipinski’s rule with machine learning are used for the virtual screening of many compounds from new drug discovery. The Gradient boosting Classifier is found to be the most suitable algorithm for predicting drug approval with the geometric mean, F1-score, and AUC of 0.95, 0.96, and 0.98, while the Random Forest Classifier is the best for drug target interactive with accuracy, F-1 score, and AUC of 0.75, 0.74, and 0.80, respectively. Moreover, we deploy machine learning with a web application and reduce time and cost in finding new drugs, increasing the chances of success in finding new treatments to reduce the mortality rate of breast cancer patients.