End-to-end de novo design of Zn²⁺ metallohydrolase binders: an open-source canonical pipeline anchored by LigandMPNN's metal-coordination recovery

Correction (new version, 2026-08-18) — the §3.6 zinc-binding-group sentence reported a base rate, not a result. Corrects the version of 2026-07-18. The statement that the most reproducible candidates are dominated by sulfonamide and hydroxamate zinc-binding groups, "the correct catalytic-Zn²⁺ pharmacophores for an MMP-1 campaign", is withdrawn from the abstract and §3.6. The library was filtered on zinc-binding-group presence before any energy was computed, so that composition is the base rate of the library rather than anything σ_E selected: measured against it, the low-σ_E set shows no enrichment in zinc-binding groups (odds ratio 0.90, p = 0.034, n = 13,158 as of the 2026-08-13 library snapshot; 0.89 on the 14,284 scored by 2026-08-18) and is depleted roughly threefold in hydroxamates (odds ratio 0.32, p = 4 × 10⁻²¹). The closing sentence of §3.6, which offered σ_E as a prioritization gate before downstream commitment, is narrowed to match the companion records, whose 2026-08-12 versions withdrew σ_E as a binder filter. The 1HFC recovery benchmark, the ESM-C oracle, the cost table, the HETATM silent-fallback finding and the MAP-Elites archive statistics are unaffected.

Correction (new version, 2026-07-18) — integrity correction. Corrected version (2026-07-18), a fabricated-panel identity misstatement removed. A primary-source audit (2026-07-16) established that the referenced calibration panel's annotations are fabricated. This version removes the false statement that 'CHEMBL406 [is] a known MMP-1 inhibitor in the ChEMBL database' and reframes all references to CHEMBL406 and to the '15 ChEMBL MMP-1 ligands' as calibration-panel structures of unverified identity. The headline result (LigandMPNN roughly doubles Zn-coordinating-residue recovery on 1HFC, 95.3% vs 46.4%, with ESM-C corroboration), the de novo library and MAP-Elites statistics, and the HETATM silent-fallback finding are all unaffected — none use the panel. Versioned correction, not a retraction. In silico only.

De novo design of binders against Zn²⁺ metallohydrolases (matrix metalloproteinases, carbonic anhydrases, thermolysins, and related catalytic-metal enzymes) remains one of the most demanding stress tests for modern generative protein modeling. The catalytic geometry of these enzymes depends on a small set of coordinating residues (typically His/Asp/Glu/Cys) whose identity and side-chain rotamer states must be preserved through every stage of an end-to-end design pipeline. We present a fully open-source canonical pipeline that integrates four publicly released components — RFdiffusion3 for backbone generation, LigandMPNN for metal-aware inverse folding, FlowPacker for side-chain refinement, and AlphaFold3 / Boltz-2x / Chai-1 for cofold validation — into a reproducible workflow we apply to the matrix metalloproteinase-1 (MMP-1) catalytic domain. The pivotal stage is sequence design: on the 1HFC reference scaffold (157 residues, 2× Zn²⁺ + 1× Ca²⁺), LigandMPNN recovers 95.3% of the six Zn-coordinating positions versus 46.4% for plain ProteinMPNN. The disparity is most pronounced at the structural-Zn triad (His183/Asp185/His196), where ProteinMPNN scores 0% versus LigandMPNN's 90.6%. An orthogonal ESM-C 600M zero-shot likelihood oracle independently confirms that LigandMPNN sequences are more native-like (mean perplexity 2.85 vs 3.03). We document a silent failure mode — when HETATM lines are stripped during preprocessing, LigandMPNN reports use_ligand_context=True but quietly degenerates to ProteinMPNN behavior — and provide a preflight check. The pipeline composes naturally with neural network potential (NNP) ranking (paper_A) and physicality-steered cofold validation (paper_B, --use_potentials). Extending the pipeline with a generative front-end, we further assemble an in-silico de novo MMP-1 candidate library of 2,861 molecules and rank it with a GFN2-xTB energy-reproducibility (σ_E) matrix (3,968 cofold-pose energy shards across 246 method×solvent cells); 556 candidates reach σ_E < 0.5 kcal/mol. Measured against the base rate of the ZBG-filtered library they are drawn from, these reproducible candidates carry no enrichment for zinc-binding groups (odds ratio 0.90) and are depleted roughly threefold in hydroxamates (odds ratio 0.32, p = 4 × 10⁻²¹), so σ_E orders the library by energetic reproducibility and not by pharmacophore class. We argue that this open canonical stack now matches or exceeds the design quality of closed alternatives (AlphaProteo) at zero licensing cost for academic users.

Keywords: de novo enzyme design, metalloenzyme, matrix metalloproteinase, LigandMPNN, RFdiffusion, FlowPacker, AlphaFold3, Boltz-2x, open source, reproducibility.