Search for a command to run...
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 Stability v3 Engineering Constraints (Locked Diagnostic Failures) Four systems were run and deliberately retained as locked failures. These are not discarded results — they are documented proof that the methodology correctly identifies when a computational model is physically insufficient. ID System Failure Interpretation ENG-001 FePorphyrin_FeIII bare cation Sz collapse Bare Fe³⁺ cation without ligand environment is unphysical ENG-002 FePorphyrin_FeII high-spin quintet Off-target spin Quintet state inaccessible in minimal fragment ENG-003 FePorphyrin_FeIII thiolate Spin escalation ~5 Ha gap Full porphyrin macrocycle required for Fe³⁺ ENG-004 Mn_SOD_minimal Spin escalation Sz=4.5, ~20 Ha gap Mn²⁺ high-spin requires extended coordination shell What σ Means σ is the standard deviation of the best VQE energy found across five independent simulation runs, converted to kcal/mol. It measures how reproducibly the optimizer finds the same energy basin when started from different random initializations. A low σ means the electronic landscape has a single well-defined minimum — the system always ends up in the same place regardless of where it starts. This is Rigid Stability, and it means the model is reliable enough to base decisions on. A high σ means the optimizer finds different energy basins on different runs — the landscape is rugged and the electronic state is sensitive to initial conditions. This is Coherent Open-Shell, and it means results should be interpreted more carefully. The σ value is not the only criterion. A result must also pass the sector audit (⟨N⟩ and ⟨Sz⟩ both within ±0.1 of their targets, dom_p ≥ 0.99) on every seed before it is locked. Computational Details Parameter Value Method Penalized VQE, UCCSD-like ansatz (depth=6) Basis set LANL2DZ (ECP on transition metals) Qubits 16 (chelator systems) or 20 (metalloenzyme systems) N-sector penalty λ_N = 2.0 Ha Sz-sector penalty λ_Sz = 5.0 Ha Seeds per system 5 (9 for Zn_CA2_minimal) Hardware Single NVIDIA L40S GPU, 48 GB VRAM Framework PyTorch 2.3 / TorchQuantum 0.1.8 / OpenFermion 1.6 / PySCF 2.5 Associated Resources Manuscript under review: ct-2026-00566z — Journal of Chemical Theory and Computation Platform: https://www.quantum-clarity.com/drug-discovery v1 dataset: https://doi.org/10.5281/zenodo.19142883 v2 dataset: https://doi.org/10.5281/zenodo.19163956