Unlocking Drug Combinations: A Novel Graph-Based Approach for Predicting Anti-Cancer Synergy

In the quest for effective cancer therapies, exploring combinatorial drug regimens is crucialto harness synergistic interactions and overcome drug resistance. However, identifying synergistic drugpairs faces challenges due to the vast combinatorial space and experimental limitations. This studyintroduces ClusterSyn, a novel machine learning-powered framework for classifying anti-cancer drugsynergy scores. Utilizing drug synergy scores on cancer cell lines, ClusterSyn employs a two-stepapproach involving drug clustering and synergy score prediction using a fully connected deep neuralnetwork. For each cell line in the training dataset, a drug graph is constructed, with nodes representingdrugs and edge weights denoting synergy scores between drug pairs. Drugs are clustered using theMarkov clustering (MCL) algorithm, aligning similar synergy profiles. Subsequently, vectorsrepresenting the similarity of drug pairs to each cluster are input into the deep neural network forsynergy score prediction (synergy or antagonism). Comparative analysis with clustering and regressionbasedmethods, including DeepSynergy and DeepDDS, demonstrates the superior performance ofClusterSyn on diverse datasets such as Oniel and Almanac. These results underscore the remarkablepotential of ClusterSyn as a versatile tool for predicting anti-cancer drug synergy scores.

کلیدواژه ها:

Drug synergy ، Clustering ، Prediction ، Machine learning