An atom-bond transformer-based message passing neural network for molecular property prediction.

Predicting properties of small molecules using MPNN on QM9 dataset

Official Implementation of Multi-Masked Aggregators for Graph Neural Networks in Pytorch and PyTorch Geometric

Multiscale exploration of informative latent features for accurate deep eutectic solvents viscosity prediction

Graph Neural Network creation module, implemented in Tensorflow 2 with examples using the module and the iGNNition library for fast GNN prototyping.

A code library for deep reinforcement learning with relational data

Melty Molecule is a molecular property prediction system that leverages graph neural networks to predict melting points of chemical compounds. The model uses a hybrid architecture combining Principal Neighborhood Aggregation (PNA) and Graph Isomorphism Networks (GIN) with edge updates and virtual node connectivity.

This project implements mechanisms to mitigate over-squashing, improving GNN performance on tasks requiring deep Networks.

Scaffold-split benchmark suite for ADMET property prediction (hERG, solubility, CYP3A4, Caco-2, clearance) with Morgan FP+RF and MPNN baselines, plus ECE/calibration metrics for uncertainty-aware evaluation. PyTorch · RDKit · DeepChem · scikit-learn.

Exploring numerous natural language processing techniques for antimicrobial peptide sequences.

A Continuous Kernel Message Passing Neural Network (MPNN) utilizing Dynamic Filter Generation to explicitly model continuous interatomic distances. Achieves accuracy (R^2 > 0.9) on the OQMD database for predicting quantum properties, offering a 10^6x speedup over DFT simulations.

A Message-passing neural network (MPNN) for molecular property prediction