Seleksi Fitur Molekuler untuk Memprediksi Sinergitas Obat pada Sel Kanker

Abstract

Identifying synergistic drug combinationsin cancer treatmentis challengingdue to the complex molecular circuitry of cancer and the exponentially increasing number of drugs. Therefore, computational approaches forpredicting drug synergy are crucialin guidingexperimental effortstowardfinding rational combination therapies. This research selects the molecular features of cancer cells with a diffusion network-based approach.Additionally, a modelis developed using non-linear regression algorithms,namely Random Forest, Extremely Randomized Tree, and XGBoost, to predict the synergy score of drug combinations against the selected cancer cell features.Thedata used are118 drug combination screening data and 85 cancer cell molecules provided by AstraZeneca-Sanger DREAM Challenge. Thepredictionresults indicate that as the data size increases, the correlation value of the model improves, leading to better prediction accuracy. The influential feature analysis revealed that the three most influential mutation features in the AKT_1 and PIK3C drug combination model wereATP8B3, ERBB2, and RNF8. In the drug combination model BCL2_BCL2L1 and FGFR, the three mostinfluential mutation features wereBACH1, ODF2, and BFAR. In the MAP2K_1 and PIK3C drug combination model, TP53, IL12p40*, and SOX4 werethe most influential features.All of these features have a connection between the mutation features and cell lines, aligning with the therapeutic targets of the three-drug combinations, which were the focus of this study.