WADDAICA

WADDAICA integrates deep learning and classical algorithms to design and modify novel drug compounds, enabling scaffold hopping and binding affinity prediction for drug discovery.

Key Features:

• Dual-Module Architecture: Structured into a Deep Learning Module and a Classical Algorithms Module to provide complementary approaches for compound design.
• Deep Learning Module: Utilizes advanced deep learning techniques to facilitate scaffold hopping and scoring, with validation reported against the PDBbind database.
• Classical Algorithms Module: Implements traditional computational methods for systematic exploration and modification of molecular structures.
• Performance Metrics: Demonstrates higher Pearson and Spearman correlation coefficients for binding affinity prediction compared to AutoDock Vina.

Scientific Applications:

• Scaffold Hopping: Identifying novel chemical scaffolds by altering existing structures to expand chemical space for drug candidates.
• Drug Optimization: Refining lead compounds to improve properties such as potency, selectivity, or pharmacokinetics.
• Binding Affinity Prediction: Predicting compound–target binding strengths to prioritize candidates during early-stage drug development.

Methodology:

Deep learning models trained on datasets of known drugs are used for generation and scoring (with validation against PDBbind), and classical algorithms are employed for systematic exploration and rule-based modification of molecular structures.