Graphlet-MP: Graphlet Histogram Representation Database of Inorganic Crystals

Graphlet-MP is a precomputed database of graphlet histogram representations for 149,082 inorganic crystals from the Materials Project. Each material is described by 79 distributions spanning three hierarchical graphlet orders extracted from its crystallographic information file via screened Voronoi tessellation: 10 first-order distributions over atomic sites, 21 second-order distributions over bonded pairs, and 48 third-order distributions over bond-angle triplets. Each material is provided as a graphlet JSON file holding the enumerated motifs and their multiplicities, and a histogram JSON file holding the 79 distributions binned on a shared 20-bin scheme, so that all materials occupy a common fixed-dimensional feature space suitable for Earth Mover's Distance comparison and kernel-based property prediction. Structures were retrieved from the Materials Project (v2025.09.25) on 2026-04-06. Of the 154,879 source CIFs processed, 5,793 were excluded on featurization failure (5,621 with sub-ångström interatomic contacts, 170 missing an elemental attribute, and 2 with no Voronoi neighbours), and 4 legacy mvc- entries were omitted so that every record carries a standard mp- identifier. The archive also includes the shared bin definitions and elemental attribute tables needed to extend the database to new structures. Full documentation is provided in README_DATASET.md.