ShEPhERD-2 Supplementary Data
Abstract
Supporting data for ShEPhERD-2, organized into three archives containing MOSES-aq generation and evaluation outputs, MolGenBench samples and evaluation results, and associated case study data. When extracted, the archives form a common shepherd2-moses/ directory. The uncompressed dataset is approximately 8.8 GB and contains 180,604 files: mosesaq is 309 MB (830 files), molgenbench is 5.2 GB (105,109 files), and the case_studies is 3.3 GB (74,664 files). The data necessary to re-run experiments in the paper are archived in shepherd2-moses-input_data.zip, containing relevant conformers, interaction profiles, and necessary docking .pdbqts. Please view the README in this file for more detail. The training data for ShEPhERD-2 is also provided as HDF5 files: moses_aq_hdf5. This consists of 5 files for a total of 2.1GB.