Advancing mutagenicity predictions in drug discovery with an explainable few-shot deep learning framework
Luis H.M. Torres, Sofia Marques da Silva, Joel Perdiz Arrais, Catarina Pimentel, Bernardete Ribeiro · Digital Discovery · 2025
Meta-GTMP is a few-shot GNN-Transformer framework that predicts Ames mutagenicity by learning from strain-specific tasks. It integrates graph embeddings and self-attention using a meta-learning strategy to generalize from limited data.