Data

DeePEST-OS: A Generic Machine Learning Potential for Accelerating Transition State Search in Organic Synthesis Dataset DOI: 10.5061/dryad.2z34tmpzg Description of the data and file structure This reaction database is generated by manuscript“DeePEST-OS: A Generic Machine Learning Potential for Accelerating Transition State Search in Organic Synthesis”. Files and variables File: DORTS-9K.h5 Description: contains 9,017 reaction data for the DeePEST-OS model training. The IRC folder includes the geometric structures of five key IRC path points (ts, reactant, product, F_rev_max, F_forw_max), computed via GFN2-xTB, with molecular energies and atomic forces calculated at the wB97M-V/def2-TZVP level. In the NMS, 20 conformers per point were generated using quantum normal modes, and their molecular energies/atomic forces were computed at the wB97M-V/def2-TZVP level. The DeePEST-OS model was trained on all NMS data.User can read the data using the script read_DORT-9K.py. │├── IRC/│ ├── ts/│ │ ├── rxn1│ │ │ ├──rxn_name│ │ │ ├──symbols│ │ │ ├──coordinates│ │ │ ├──energy│ │ │ └──forces│ │ ├── rxn2│ │ └── ...│ ├── reactant/│ │ ├── rxn1│ │ └── ...│ ├── product/│ │ └── ...│ ├── F_forw/│ │ └── ...│ └── F_rev/│ └── ...│└── NMS/ ├── rxn1 ├── rxn2 └── ... File: DORTS.tar.gz Description: contains 65,116 reaction data beyond DORTS-9K. The IRC folder includes the geometric structures of five key IRC path points (ts, reactant, product, F_rev_max, F_forw_max) computed via GFN2-xTB. In the NMS, 20 conformers per point were generated using quantum thermal normal mode sampling. Notably, DORTS.tar.gz only contains geometric structures, without directly providing high-precision molecular energies and atomic force matrices.