About the NR-Predictor

A comprehensive deep learning framework for nuclear receptor ligand binding prediction and functional classification.

Model Framework

Module A: 12 single-task ML models using molecular fingerprints and descriptors for individual NR target predictions.
Module B: Message-passing neural network with molecule-level attention mechanism for graph-based molecular representation learning.
Module C: Meta-learner with task-specific attention that combines outputs from Modules A and B for final binding prediction.
Module D: Functional classifier integrating molecular descriptors with Vina-derived interaction features for downstream functional classification.
Attention Mechanism: K-layer GRU-based attention with super node aggregation for interpretable predictions.

Laboratory of Molecular Modeling and Design
East China University of Science and Technology