Drug synergy model for malignant diseases using deep learning

Abstract

Drug synergy has emerged as a viable treatment option for malignancy. Drug synergy reduces toxicity, improves therapeutic efficacy, and overcomes drug resistance when compared to single-drug doses. Thus, it has attained significant interest from academics and pharmaceutical organizations. Due to the enormous combinatorial search space, it is impossible to experimentally validate every conceivable combination for synergistic interaction. Due to advancement in artificial intelligence, the computational techniques are being utilized to identify synergistic drug combinations, whereas prior literature has focused on treating certain malignancies. As a result, high-order drug combinations have been given little consideration. Here, DrugSymby, a novel deep-learning model is proposed for predicting drug combinations. To achieve this objective, the data is collected from datasets that include information on anti-cancer drugs, gene expression profiles of malignant cell lines, and screening data against a wide range of malignant cell lines. The proposed model was developed using this data and achieved high performance with f1-score of 0.98, recall of 0.99, and precision of 0.98. The evaluation results of DrugSymby model utilizing drug combination screening data from the NCI-ALMANAC screening dataset indicate drug combination prediction is effective. The proposed model will be used to determine the most successful synergistic drug combinations, and also increase the possibilities of exploring new drug combinations.