GDF5 Trp414/Trp417 Structural Models and Reproducibility Data

This dataset accompanies the manuscript “In Silico Structural Analysis of the GDF5 Mature Domain Identifies a Putative Conserved G-W-(X/–)-X-W-I/V Motif Containing Invariant Trp414 and Trp417” by Michael Picker, submitted to PROTEINS: Structure, Function, and Bioinformatics. The deposit contains two complementary computational structural datasets. Part 1 — GDF5 Wild-Type, Trp414Arg, and Trp417Ser AlphaFold-Multimer Structural Ensemble ZENODO_PART1_RELEASE_4_0_1.zip contains 45 unrelaxed AlphaFold2-Multimer v3 predictions of the GDF5 mature-domain homodimer, comprising 15 predictions each for wild-type GDF5, p.Trp414Arg, and p.Trp417Ser. Each genotype includes five model parameterizations and three seeds. The archive includes PDB coordinates, model assignments, model-confidence and structural-consistency analyses, analysis code, and validation records. Part 2 — GDF5 Wild-Type and Substitution Models Derived from Experimental Structures 1WAQ and 8BWL ZENODO_PART2_RELEASE_4_0_1.zip contains ten computational models prepared using previously published experimental GDF5 structures 1WAQ and 8BWL as templates. For each template, the dataset includes wild-type GDF5 and the p.Trp417Ser, p.Trp417Phe, p.Trp417Arg, and p.Trp414Arg substitutions. The archive includes ten pre-relaxation input structures, ten final structures following restrained energy minimization, modeling provenance, quality-control records, and measurements of minimum side-chain heavy-atom distances between residues corresponding to GDF5 positions 414 and 417. The original experimental structures were determined by their respective depositors. This deposit provides computational predictions and template-derived models, with attribution to the source structures. The models and geometric measurements support structural comparisons and hypothesis generation. Functional effects require experimental investigation. Version 4.0.1 includes file-level SHA-256 manifests and release documentation. Reanalysis of the archived AlphaFold models was numerically checked, and the ten reported side-chain distances were independently recalculated from the final coordinates. Selected reanalysis and figure workflows were tested on macOS in an isolated Python environment using these datasets together with the companion code package. Upstream ColabFold prediction and historical OpenMM minimization were outside that execution test. Original research data and documentation are licensed under CC BY 4.0. Original software is licensed under MIT, as specified in the archives. Source PDB archive data are provided under CC0, with source identifiers and attribution retained.