RNA pocket atlas, version 1.0: RNA-small molecule binding pockets with structural, compositional and morphometric descriptors

Data and code of the article "SharpShape: a curated atlas and web server linking the shape of RNA-small molecule binding pockets to their structure and ligands" (Ziani L, Badel A, Pasquali S, Flatters D, Camproux A-C, manuscript submitted, 2026). The atlas describes 820 RNA-small molecule binding pockets extracted with a single protocol from 455 RNA-only complexes of the Protein Data Bank (PDB and ligand identifiers obtained from HARIBOSS as retrieved in June 2026). A pocket contains the complete RNA nucleotides with at least one heavy atom within 6 Å of a ligand heavy atom. The reference subset contains 278 pockets: within each of the 153 RNA sequence clusters (MMseqs2, 90% identity), all chains pooled, a pocket is kept only if its ligand is chemically distinct (Tanimoto < 0.70) from those already kept, in a fixed selection order (seed 2026). Each pocket is described by its size, shape (eigenvalue ratios and shape archetype), base pairs, stacking, backbone links, secondary- and tertiary-structure features, sugar conformation, nucleotide composition, ligand properties, ligand burial and RNA class. Files: pocket table of the full atlas (820 pockets, reference subset given by the column in_reference), pocket table of the reference subset (278 pockets), descriptor dictionary, ligands, RNA chains, the 533 RNA-only complexes listed in HARIBOSS with their processing status, the 111 pockets containing modified nucleotides, pocket coordinates (PDB format), version 1.0.0 of the command-line tool (pocket extraction, descriptors, shape archetypes) and the scripts of the main statistical analyses of the article. See README.md for the content of each file. Web server: https://sharpshape.rpbs.univ-paris-diderot.fr (same tables and pocket files). Command-line tool: https://github.com/ZianiL/sharpshape