GeoScatt-GNN: A Geometric Scattering Transform-Based Graph Neural Network Model for Ames Mutagenicity Prediction

TL;DR

This paper introduces a novel graph neural network model, GeoScatt-GNN, that combines geometric scattering transforms with GNNs to improve mutagenicity prediction accuracy, demonstrating significant advancements over traditional methods.

Contribution

It presents a new GNN architecture integrating geometric scattering features, outperforming existing models in mutagenicity prediction tasks.

Findings

2D scattering coefficients outperform traditional descriptors

Hybrid GGS and GIN approach achieves high accuracy

GGS-enhanced GNN significantly improves prediction results

Abstract

This paper tackles the pressing challenge of mutagenicity prediction by introducing three ground-breaking approaches. First, it showcases the superior performance of 2D scattering coefficients extracted from molecular images, compared to traditional molecular descriptors. Second, it presents a hybrid approach that combines geometric graph scattering (GGS), Graph Isomorphism Networks (GIN), and machine learning models, achieving strong results in mutagenicity prediction. Third, it introduces a novel graph neural network architecture, MOLG3-SAGE, which integrates GGS node features into a fully connected graph structure, delivering outstanding predictive accuracy. Experimental results on the ZINC dataset demonstrate significant improvements, emphasizing the effectiveness of blending 2D and geometric scattering techniques with graph neural networks. This study illustrates the potential of GNNs and GGS for mutagenicity prediction, with broad implications for drug discovery and chemical safety assessment.