Supporting data for "Protein language model-assisted enzyme mining and experimental characterization of bacterial GH28 polygalacturonases for pectin depolymerization"

Raw measurements, computational inputs and outputs, and analysis code underlying every figure and table of the associated manuscript. Four bacterial glycoside hydrolase family 28 (GH28) polygalacturonases are characterized: FLCH_PG (Flavobacterium cheongpyeongense), FLHY_PG (F. hydrophilum), FLAG_PG (F. agrisoli) and PAMA_PG (Paenibacillus massiliensis), together with their action on polygalacturonic acid and on commercial low- and high-methoxyl pectin. Contents 01_kinetics — per-replicate absorbances with matched enzyme-free blanks, the D-galacturonic acid calibration series, and the single generator script that refits every parameter in Table 1 at runtime from the raw CSV (no hard-coded values), reproducing Table 1, Fig. 3 and Fig. S5. 02_biochemical_profiles — temperature and pH optima, thermostability time-courses (50/60/70 °C x 10-120 min), and metal-ion and EDTA effects. 03_pectin_hpsec — high-performance size-exclusion chromatography of pectin depolymerization alone and with Bacillus licheniformis pectin methylesterase, with molecular-weight and peak-area calculation. 04_sequence_mining — protein language model-assisted candidate prioritization, sequence similarity network and MCL clustering, and the activity screen of 16 bacterial candidates against 3 fungal references. 05_structure_docking — BioEmu apo ensembles (ten representative conformers per enzyme), docking inputs and poses, quality-pose selection, contact analyses and the numbered pipeline scripts. 06_manuscript_tables — the published table values for cross-checking against the raw data. Naming note. In the raw data files, scripts and dictionary keys, PAMA_PG appears under its original internal key TEHA_PG. The two refer to the same enzyme; the internal key is preserved so that data joins remain reproducible. Interpretation limit for the structural data. A 1KCD self-redock positive control reproduced the crystal ligand conformation but displaced its catalytic register by more than 2 Å, all three catalytic aspartates were modelled as neutral at pH 5.0, and pose selection was conditioned on catalytic-aspartate proximity. These calculations support only the steric compatibility of oligogalacturonate with a conserved groove; they do not resolve the scissile-bond register, interaction energetics or subsite architecture. The reported ensemble variability is positional dispersion across ten aligned conformers and is not a molecular-dynamics time-series RMSF. All kinetic constants are apparent and mass-based, because polygalacturonic acid is a polydisperse polymer.