D3CARP

D3CARP predicts drug-target interactions and supports virtual screening by integrating multiple-conformation ensemble docking, 2D/3D ligand similarity searches, and deep learning to prioritize compound–target pairs for therapeutic investigation.

Key Features:

• Multiple-Conformation Based Docking: Molecular docking using 9,352 conformations across 1,970 targets with positive controls and ensemble docking approaches, reported docking accuracy of 0.44.
• Ligand Similarity Search: 2D and 3D ligand similarity searches across approximately 2 million target–ligand pairs using five different search methods with an average accuracy of 0.94.
• Deep Learning Approaches: Deep learning models for DTI prediction with reported accuracy of 0.89.
• Integration of Positive Controls and Disease Annotation: Inclusion of positive compounds as references and annotation of related diseases for therapeutic targets to support disease-based DTI analyses.
• Cross-Validation and Broad Applicability: Emphasis on cross-validation of predicted outcomes and application across diverse disease types using multiple computational methods.

Scientific Applications:

• Target Prediction: Prioritizes potential drug–target interactions across 1,970 targets using docking, similarity, and deep learning evidence.
• Virtual Screening: Screens approximately 2 million target–ligand pairs using ensemble docking and ligand similarity to identify candidate compounds.
• Drug–Target Interaction (DTI) Studies: Supports DTI studies with integrated positive controls and disease annotations to relate targets to therapeutic contexts.
• Model Evaluation and Validation: Enables assessment of prediction robustness through cross-validation and reported accuracy metrics for docking, similarity searches, and deep learning.

Methodology:

Molecular docking with ensemble conformations (9,352 conformations across 1,970 targets), 2D and 3D ligand similarity searches across ~2 million target–ligand pairs using five search methods, deep learning models for DTI prediction, use of positive controls and disease annotations, and cross-validation of predictions.