SynPred

SynPred predicts synergistic effects of anticancer drug combinations by integrating multiomics and biophysical data with ensembles of machine learning and deep learning algorithms to inform drug discovery and personalized cancer therapy.

Key Features:

• AI-Driven Predictive Modeling: Employs ensembles of machine learning (ML) and deep learning (DL) algorithms to link omics and biophysical traits and predict anticancer drug synergy.
• Multiomics and High-Throughput Data Integration: Integrates multiomics datasets generated from high-throughput screening across populations and cell types, including genomic data and biophysical traits.
• Input Representation: Uses SMILES (Simplified Molecular Input Line Entry System) representations of two drugs as molecular input for combination predictions.
• Performance Metrics: Reports performance on independent test sets with accuracy 0.85, precision 0.77, recall 0.75, AUROC 0.82, and F1-score 0.76.
• Data Interpretability: Implements feature importance approaches to provide interpretable predictors of synergistic interactions.

Scientific Applications:

• Drug Discovery and Development: Predicts effective anticancer drug combinations to accelerate identification of synergistic therapeutic strategies.
• Personalized Medicine: Supports tailoring combination therapies to individual patient profiles by integrating diverse biological data to address cancer heterogeneity and genomic variability.

Methodology:

Integrates multiomics datasets from high-throughput screening with ensembles of ML and DL algorithms, uses SMILES inputs for drug representation, and applies feature importance methods to predict synergistic drug interactions.