New publication in Nature communications - TopEC: a new AI-based method for the prediction of Enzyme Commission classes

A new study was published in the Nature communications by the Gohlke group in collaboration with Helmholtz AI. This study presents TopEC, a new 3D graph neural network based prediction tool to predict Enzyme Commission (EC) classes. In comparison to older existing methods such as EnzyNet or DeepFRI, TopEC significantly improves EC classification prediction (F-score: 0.72). Trained on a balanced set of experimental and computationally generated enzyme structures covering a vast functional space (>800 ECs), the method offers a robust predictability without fold bias at residue and atomic resolutions. In contrast to pure sequence-based state-of-the-art methods such as CLEAN, TopEC is based on local 3D descriptors and atomistic graphs, thus allowing the user to probe the significance of individual atoms or residues in the prediction results. We expect TopEC to be useful in enzyme discovery, specifically for cases of homologous enzymes sharing both similar sequences and folds, but not the same catalytic function.

TopEC is available as a repository on GitHub: https://github.com/IBG4-CBCLab/TopEC and https://doi.org/10.25838/d5p-66.

The publication is available here: https://www.nature.com/articles/s41467-025-57324-5