SYNDEEP: a deep learning approach for the prediction of cancer drugs synergy

Area of Science:

• Computational biology
• Bioinformatics
• Cancer pharmacology

Background:

• Drug combinations are crucial for effective cancer therapy, but experimental identification is resource-intensive.
• Exploring all possible drug combinations experimentally is often infeasible due to cost and time constraints.
• Developing computational methods is essential to accelerate the discovery of synergistic drug combinations.

Purpose of the Study:

• To present a novel computational approach for predicting synergistic drug combinations in cancer therapy.
• To develop and validate a deep neural network model for synergy prediction.
• To assess the model's performance against traditional machine learning algorithms.

Main Methods:

• Utilized a deep neural network (DNN)-based binary classification model.
• Integrated diverse data types including physicochemical, genomic, protein-protein interaction, and protein-metabolite interaction features.
• Compared DNN performance with shallow neural network (SNN), k-nearest neighbors (KNN), random forest (RF), support vector machines (SVMs), and gradient boosting classifiers (GBC).

Main Results:

• The proposed DNN model achieved high accuracy (92.21%) and AUC (97.32%) in tenfold cross-validation.
• The DNN model outperformed other evaluated machine learning classifiers in predicting synergistic drug combinations.
• Integration of physicochemical and genomics features significantly improved prediction accuracy for drug synergy.

Conclusions:

• Deep neural networks offer a powerful and accurate computational approach for predicting synergistic drug combinations.
• The developed model provides a valuable tool to expedite the identification of effective combination therapies in cancer treatment.
• Combining multiple feature types enhances the predictive capability for cancer drug synergy.