This dataset consists of 1,000 coordinate files (in the CHARMM psf/cor format) for the QM/MM minimum energy pathways of the deacylation reactions between a Class A beta-lactamases (GES-5) and the imipenem antibiotic molecules. All pathway conformations were optimized at DFTB3/3OB-f/CHARMM36 level with 36 replicas. All single point calculations and charge population analysis were done at B3LYP-D3/6-31+G(d,p)/CHARMM36 level. 0.paths_ges_imi_d1.tar.gz: 500 pathway conformations for GES-5/IPM-Delta1 deacylation reactions. 0.paths_ges_imi_d2.tar.gz: 500 pathway conformations for GES-5/IPM-Delta1 deacylation reactions. 1.eners.zip: The single point replica energies along all GES-5/IPM pathways. 1.chrgs.zip: The NBO charges of the QM region of all replica conformations along all GES-5/IPM pathways. 2.datasets.zip: The Python codes to postprocess the molecular data and the featurized the NumPy arrays. 3.gnn.zip: The Python codes that implements the edge-conditioned graph convolutional NN to predict the deacylation barriers. 5.representative_conf.zip: The pathway conformations of all cluster centroids and an energetic representative (pathway id 22) pathway. Note: This file also serves as a peephole of how the pathway conformations from Reaction Path with Holonomic Constrains calculations looks like. 6.benchmark.zip: The benchmark calculations that validates the DFTB3/3OB-f/CHARMM36 against DFTB3/3OB/CHARMM36 and B3LYP/6-31G(d,p)/CHARMM36 level of theory on the energetic representative (pathway id 22) pathway conformations. p.figures.zip: A series of Jupyter Notebooks that produces the visualizations in the work. README.md: A markdown file that contains additional descriptions. environment.yml: The Conda environment used for the graph-learning. Z. Song and P. Tao, Graph-Learning Guided Mechanistic Insights into Imipenem Hydrolysis in GES Carbapenemases. Electron. Struct. 2022, 4, 034001. DOI: 10.1088/2516-1075/ac7993