RIProfile data generated from inverse mixed-solvent molecular dynamics simulations

RIProfile data generated from inverse mixed-solvent molecular dynamics simulations Residue interaction profiles (RIprofiles) and probe structures from the inverse mixed-solvent molecular dynamics (inverse MSMD) study. Contents The release is split into one self-contained ZIP per probe to keep the file count within Zenodo's 100-file-per-record limit while letting users download only the probes they need. Top-level files: qinvmsmd_profiles_v1/ ├── README.md ├── probes.csv # CID, CANONICAL_SMILES (94 rows) └── {CID}.zip # one per probe (94 archives total) Each {CID}.zip expands to: {CID}/ ├── probe_structures/{CID}.{mol2,pdb,smi} └── interaction_profile/{CID}_{RES}_profile.dx.gz (19 files) {RES} is the three-letter amino-acid code of the surrounding residue type. Profiles cover 19 residue types — all 20 standard amino acids except GLY, which has no Cβ atom and is therefore not represented in the Cβ-based density used here. ZIP archives are stored without additional compression (the .dx.gz files are already gzipped); extract with any standard unzip implementation. File contents per probe ID: {CID}.mol2 — GAFF2 atom types + AM1-BCC charges (Gaussian/AmberTools) {CID}.pdb — same conformer in PDB {CID}.smi — single-line SMILES RIprofile (.dx.gz) Each *_profile.dx.gz is a gzipped OpenDX scalar grid on a 100×100×100 cubic mesh (1 Å spacing, origin at −50 Å in each axis) defined in the probe's local frame. The stored values are the natural logarithm of the spatial density of residue Cβ atoms of the given residue type around the overlaid probe, aggregated across the 15 profile-construction proteins listed below. A small positive floor is applied to the density before the log so that no NaN/-inf values appear; observed values lie in roughly [−18.2, 4.7]. Because the grid is in the probe's frame, RIprofiles are independent of any target protein. Higher values indicate residues of that type are more favored at that position relative to the probe. Decompress with gunzip (or read directly via gzip in code) and parse with any OpenDX-compatible tool (e.g. MDAnalysis.coordinates.DX, PyMOL load, GridDataFormats). Probe naming The CID (e.g. A17, E14) is the probe identifier from our internal cosolvent library, also used in the parent project (https://github.com/keisuke-yanagisawa/inverse_msmd). EZZ is a manually flipped amide conformer of E17 (same SMILES, different 3D conformer); its EZZ.mol2 was synthesized by transferring EZZ.pdb coordinates onto the E17.mol2 template (atom names match exactly). Profile-construction proteins The 15 proteins used to construct the RIprofiles (see Methods of the paper) are public PDB entries: 1BK9, 1CXV, 1E0X, 1H4G, 1H60, 1HEE, 1JZF, 1TT1, 1TU6, 1W4P, 1WBI, 1YMS, 1ZUA, 2CYB, 2WEA. License & citation The data are released under the same terms as the manuscript. If you use these profiles in your work, please cite the paper above. Reproduction The full pipeline (probe preparation → MSMD simulations → RIprofile construction → matching score) is implemented in inverse_msmd. Counts at a glance Quantity Value Unique probe CIDs 94 Per-probe ZIP archives 94 Probe structure files (across all ZIPs) 282 (94 × 3 formats) RIprofile files (across all ZIPs) 1786 (94 × 19 residues)