Enhancing interpretability in metabolomics: ranking metabolites by their impact on graph neural network predictions
A new method where Formula Difference Networks are used as inputs to predictive Graph Neural Network models, and it is found that there was a clear bias for assigning higher importance to nodes in connected subgraphs.
Francisco Traquete, C. Cordeiro, Marta Sousa Silva et al.
· BMC Bioinformatics · 0 citations