Abstract
Correction notice (November 3, 2026) The chemistry interpretations in this dataset are withdrawn. The data files are retained unchanged as an accurate record of what the software produced. The full analysis, evidence and corrected results are in ELSD v4.0, 10.5281/zenodo.23933642. Plain Language Summary This dataset reported differences between drug targets, measured from how much repeated simulations scattered. A later, stricter check measured each saved simulation directly. It found that none of them had moved past the starting approximation, so the scatter did not come from the chemistry. It came from a constant term added by a defect inside the software. The conclusions drawn from that scatter therefore do not hold. The defect has since been found and corrected, and the corrected software has been tested on a zinc enzyme model where the exact answer is known. What was found All saved states are the Hartree–Fock determinant. Every deposited statevector is the HF determinant of its target (N, S_z) sector, to within 1.3 × 10⁻⁵ Ha (v4.0 §3.1). The reported σ values are not physical-energy scatter. They are the seed-to-seed scatter of a constant penalty offset, 3.75–6.25 Ha, caused by an incompletely expanded S_z penalty (v4.0 §3.2). Root cause. The ansatz's "UCCSD-like" double-excitation block changes particle number, so it cannot move the state within the target sector (v4.0 §4, defect D2). The "UCCSD-like" label in this dataset is inaccurate. Withdrawn from this dataset the VAP-1/AOC3 "Nitrogen Rule" the DOTA vs NOTA chelator ranking (also not statistically resolved on the published numbers: F-test p = 0.83 with 5 seeds each) all σ-based interpretations carried over from v1.0 and v2.0, including the Cu²⁺ SOD 9.2× apo/bound ratio, the three-part Cu mechanistic decomposition, the Zn²⁺ CA2 warhead ladder, the Rigid Stability / Coherent Open-Shell classifications and the IDE/NEP catalytic-state maps Still valid The deposited files, as a record of the March 2026 engine's output. The Zn_squareplanar determinant-switching observation (v4.0 §3.4). How to cite The σ values in this dataset should be cited, if at all, as "seed-ensemble scatter of a penalty-term offset on Hartree–Fock-locked states (Zenodo 10.5281/zenodo.23933642, v4.0)", not as measures of electronic-landscape stability. Original description (March 25, 2026), retained unchanged below for the record. Prometheus Platform — Electronic Landscape Stability Dataset Version 3 · Quantum-Clarity LLC · March 2026 Concept DOI: 10.5281/zenodo.19142882v1 DOI: 10.5281/zenodo.19142883v2 DOI: 10.5281/zenodo.19163956Associated manuscript: ct-2026-00566z (Journal of Chemical Theory and Computation, under review) Plain Language Summary What we did We built a computational instrument that stress-tests the electronic state of drug targets and drug delivery systems — and ran it on 25 different biological systems spanning three therapeutic areas: neuroinflammation, metabolic disease, and cancer treatment. For each system, we ran five independent simulations starting from different random configurations. After every run, we verified that the simulation ended up in the physically correct electronic state — not just a low-energy state that happens to look reasonable. That verification step, which we call the sector audit, is what makes these results trustworthy. Most computational methods skip it. This version (v3) adds nine new systems to the dataset: a complete inhibitor triage series for the VAP-1/SSAO neuroinflammation target, and the first electronic stability benchmark for DOTA and NOTA — the molecular "cages" used to carry toxic payloads in cancer treatments. What we discovered Finding 1: The Nitrogen Rule for neuroinflammation drugs (VAP-1/SSAO target) VAP-1 is a copper enzyme implicated in neuroinflammation, Alzheimer's disease, and blood-brain barrier breakdown. We mapped its electronic landscape across eight different drug-target interaction states. The finding is clear and consistent across three independent inhibitor classes: drugs that use Nitrogen to interact with the VAP-1 copper center produce a stable, reproducible electronic state. Drugs that use Oxygen do not. Triazole class (N-donor): σ = 0.1747 kcal/mol — the most electronically stable Cu²⁺ result in our entire dataset Hydrazine class (N-donor): σ = 0.2003 kcal/mol — equally stable Imidazole class (N-donor): σ = 0.3235 kcal/mol — solidly stable Semicarbazide class (O-donor): σ = 1.7903 kcal/mol — ten times more electronically noisy Same protein. Same metal center. Ten times the difference in electronic stability depending on which atom does the talking. This boundary is now mapped from both sides with six independent data points across three N-donor chemotypes and one O-donor chemotype. Finding 2: The ADC chelator electronic hierarchy (cancer drug delivery) In Antibody-Drug Conjugates and radiopharmaceuticals, a chelator is the molecular cage that must hold a radioactive or toxic metal payload in a well-defined state as it travels through the bloodstream to a tumor. The industry standard is DOTA. A commonly used alternative is NOTA. We ran the first electronic stability audit on both: DOTA (12-membered ring, 4 nitrogen donors): σ = 0.4534 kcal/mol — Rigid Stability, decision-grade NOTA (9-membered ring, 3 nitrogen donors): σ = 0.5095 kcal/mol — Rigid Stability, measurably broader Both are decision-grade. But Prometheus independently ranked DOTA above NOTA in electronic stability — matching the known experimental hierarchy — using quantum geometry alone, with no prior experimental input. This demonstrates that the ELSD methodology captures a real physical signal about chelator coordination quality that is not accessible from conventional computational tools. Finding 3: The AOC3 perturbation series (mechanistic completeness) The VAP-1 active site in three additional states — true apo (no ligand), protonated donor (weakened coordination), and catalytic hydroxide — was characterized to complete the full eight-state electronic map. The T-shaped copper geometry constrains the electronic manifold so tightly that donor perturbations have less leverage than on more loosely coordinated scaffolds. Why this matters Drug discovery for metal-centered targets has a model validity problem that is rarely acknowledged: before a team ranks compounds, screens leads, or builds a medicinal chemistry program, the electronic model of the target itself needs to be trustworthy. A model that numerically converges is not necessarily in the correct electronic state. Current computational workflows do not routinely check this. They find the lowest energy state and report it. Prometheus asks a harder question: did the simulation end up in the physically correct electronic state, and does it stay there reproducibly across independent runs? For the VAP-1 neuroinflammation target, this question now has a clear answer across eight coordination states and four inhibitor chemotype classes. For the DOTA/NOTA ADC chelator problem, it has a clear answer for the first time anywhere in the literature. The practical consequence: researchers can now identify which inhibitor coordination modes produce a stable, reproducible electronic regime at the VAP-1 copper center before committing to synthesis. They can compare new chelator designs against an audited DOTA baseline before costly animal stability testing. And they can do both using a framework that explicitly documents its failures — four engineering constraints where the minimal fragment model was physically insufficient — alongside its successes. A benchmark suite containing only successes is scientifically weaker. The failures are retained because they prove the filter works. What Is New in Version 3 Version 3 adds nine new systems (90 new files) relative to v2: System σ (kcal/mol) Regime Significance Cu_AOC3_imidazole 0.3235 Rigid Stability N-donor inhibitor class 1 Cu_AOC3_semicarbazide 1.7903 Coherent Open-Shell O-donor contrast — 5.5× more rugged Cu_AOC3_minimal_apo 0.4854 Rigid Stability True apo state — no axial ligand Cu_AOC3_minimal_protonated 0.5536 Coherent Open-Shell Weakened donor perturbation Cu_AOC3_minimal_hydroxide 0.5965 Coherent Open-Shell Catalytic OH⁻ state Cu_AOC3_triazole 0.1747 Rigid Stability Tightest Cu²⁺ result in dataset Cu_AOC3_hydrazine 0.2003 Rigid Stability N-donor rule confirmed cross-class Cu_DOTA_minimal 0.4534 Rigid Stability First ADC chelator benchmark Cu_NOTA_minimal 0.5095 Rigid Stability ADC contrast — hierarchy reproduced Complete Dataset — All 25 Locked Canonical Results System σ (kcal/mol) Seeds Regime Version FePorphyrin_FeII_ls 0.3375 5 Rigid Stability v1 Zn_CA2_minimal 0.0933 9 Rigid Stability v1 Zn_CA2_imidazole 0.4367 5 Rigid Stability v1 Cu_SOD_minimal 1.7035 5 Coherent Open-Shell v1 Cu_SOD_2imidazole 0.0803 5 Rigid Stability v1 Cu_SOD_2imidazole_water 0.7419 5 Coherent Open-Shell v1 Cu_SOD_protonated 0.6402 5 Coherent Open-Shell v1 Cu_SOD_acetate 0.9017 5 Coherent Open-Shell v1 Zn_CA2_sulfonamide 0.5058 5 Rigid Stability v1 Zn_squareplanar 43.29 5 Multi-Basin (scaffold) v1 Zn_CA2_hydroxamate 0.4672 5 Rigid Stability v2 Zn_CA2_carboxylate 0.3490 5 Rigid Stability v2 Cu_SOD_minimal_CuI 0.7592 5 Intermediate/Controlled v2 Cu_SOD_minimal_water 0.9414 5 Coherent Open-Shell v2 Zn_CA2_phosphonate 0.3307 5 Rigid Stability v2 Cu_AOC3_minimal 0.5610 5 Coherent Open-Shell v2 Cu_AOC3_minimal_apo 0.4854 5 Rigid Stability v3 Cu_AOC3_minimal_protonated 0.5536 5 Coherent Open-Shell v3 Cu_AOC3_minimal_hydroxide 0.5965 5 Coherent Open-Shell v3 Cu_AOC3_imidazole 0.3235 5 Rigid Stability v3 Cu_AOC3_semicarbazide 1.7903 5 Coherent Open-Shell v3 Cu_AOC3_triazole 0.1747 5 Rigid Stability v3 Cu_AOC3_hydrazine 0.2003 5 Rigid Stability v3 Cu_DOTA_minimal 0.4534 5 Rigid Stability v3 Cu_NOTA_minimal 0.5095 5 Rigid Stabili