Decision-Grade Electronic-State
Triage for Drug Discovery
Prometheus does not rank your candidates against a model you haven't validated. It characterizes the electronic consequence of a modeled target state — and tells you whether that state is reproducible, sector-coherent, and decision-grade before you commit chemistry budget to it.
Sector-audited electronic landscape classification for metalloenzyme active sites — in days, across Fe, Zn, and Cu systems, with reproducible multi-seed validation.
FePorphyrin FeII · Zn CA2 (4 warhead classes) · Cu SOD (5 perturbations) · Cu AOC3/VAP-1 (8 states) · Cu ADC Chelators (DOTA + NOTA ladder) · Metabolic Zn Peptidases (IDE M16 + NEP M13 · multi-state series)
What we found. Why it matters.
We run electronic stress tests on drug-target interactions to find out which ones hold and which ones shake. Here are the two headline findings from our latest campaign — explained for anyone making a high-stakes decision about where to invest in drug development.
The Cancer Cage: Does the standard delivery system hold?
In cancer treatment, a chelator is the molecular "cage" that carries a toxic payload through the bloodstream to a tumor. If the cage is electronically unstable, the payload can release early — before it reaches the tumor — causing serious side effects.
We ran the first electronic stability audit on DOTA — the industry-standard cage used in Cu-based cancer drugs and radiopharmaceuticals.
The result: DOTA is electronically rigid and decision-grade. σ = 0.4534 across 5 independent tests, all clean.
We also tested NOTA — a smaller, less-established chelator. It came back measurably more rugged. σ = 0.5095 — broader, less stable, consistent with DOTA's known superiority.
Prometheus ranked them correctly — with no prior experimental input. Any new chelator design now has an audited baseline to beat before it enters costly animal testing.
The Nitrogen Rule: Which drug designs are electronically solid?
VAP-1 is a copper enzyme implicated in neuroinflammation, Alzheimer's disease, and blood-brain barrier breakdown. We mapped its electronic landscape across 8 different coordination states — catalytic variants and inhibitor classes.
The finding is clear: Nitrogen-based drug designs stabilize the target. Oxygen-based designs do not.
We tested three N-donor inhibitor classes:
→ Triazole: σ = 0.1747 — tightest result in the entire dataset. 99.99% state certainty.
→ Hydrazine: σ = 0.2003 — equally tight.
→ Imidazole: σ = 0.3235 — solidly stable.
Against the O-donor semicarbazide: σ = 1.7903 — ten times more rugged.
Same protein. Same metal center. Ten times the difference in electronic stability depending on which atom does the talking. That boundary is now mapped from both sides with six data points.
We run the wind-tunnel test. Everyone else looks at the photo.
Imagine two bridges. Both look identical from the outside. Both pass a visual inspection. But under high winds, Bridge A doesn't move an inch — while Bridge B sways violently.
Standard drug discovery tools look at the photo of the bridge. They ask: does the drug fit the target? Does the cage hold the metal in a test tube?
Quantum-Clarity runs the wind-tunnel test. We ask: is the electronic connection between drug and target reproducible, sector-coherent, and stable across independent simulations? Is the cage electronically rigid — or does it have hidden flexibility that only shows up under computational stress?
We are not replacing wet-lab stability assays. We are giving them a higher-value starting point — so teams can fail weak designs in the computer on Monday, rather than in an animal trial six months later.
The Electronic Stability Rating
This table translates the raw quantum chemistry numbers into a risk-tier framework any investment committee can read. Every σ value is derived from sector-audited, multi-seed ensemble calculations — not a single simulation, not a best guess.
| Drug / Target State | σ (kcal/mol) | Stability Rating | Risk Level | What it means |
|---|---|---|---|---|
Triazole at VAP-1N-donor · CNS pharmacophore |
0.1747 | Elite | Ultra-Low | Zero electronic wiggle. 99.99% state certainty. The tightest Cu²⁺ result in the dataset. |
Hydrazine at VAP-1N/N-donor · SSAO inhibitor class |
0.2003 | Elite | Ultra-Low | Rock-solid. Confirms N-donor stabilisation is a general rule, not imidazole-specific. |
Imidazole at VAP-1N-donor · His-mimetic inhibitor |
0.3235 | Elite | Low | Solid coordination. More stable than the natural water-bound protein state. |
DOTA chelatorADC / radiopharmaceutical gold standard |
0.4534 | Standard | Low | Industry gold standard confirmed electronically rigid. The benchmark to beat. |
NOTA chelatorADC · smaller 9-membered ring |
0.5095 | Standard | Low-Moderate | Decision-grade but measurably broader than DOTA. Consistent with known experimental hierarchy. |
Water-bound VAP-1Natural resting state |
0.5610 | Baseline | Moderate | The protein's normal flexible state. The reference point everything else is compared to. |
Semicarbazide at VAP-1O-donor · classical SSAO inhibitor |
1.7903 | Elevated Risk | High | Ten times more rugged than triazole. Same target, same metal — oxygen coordination amplifies electronic noise. |
σ = standard deviation of best energies across 5 independent seeds · all results sector-audited · public dataset: Zenodo 10.5281/zenodo.19212485
What Prometheus answers
These are not questions conventional docking or scoring workflows answer. They are upstream questions about whether the model itself is trustworthy enough to guide chemistry decisions.
Sector-audited multi-start ensemble classification tells you whether the modeled active site converges reproducibly across independent initializations — or collapses into the wrong electronic basin.
Zn CA2: Rigid Stability across 6 environments · Fe FeII ls: Rigid Stability · IDE + NEP Zn peptidase series: Rigid Stability · locked ✓For VAP-1/AOC3, N-donor inhibitors drive the Cu²⁺ manifold into Rigid Stability. For NEP, the S-donor thiolate (thiorphan class) tightens the manifold to σ=0.0969. For ADC chelators, DOTA outranks NOTA. Prometheus identifies all distinctions before synthesis.
Triazole σ=0.17 · Semicarbazide σ=1.79 · DOTA σ=0.45 · NEP thiolate σ=0.10A five-experiment SOD perturbation series spans a 21× range in landscape ruggedness while the dominant electronic family stays identical throughout all 40 runs. The NEP catalytic state map shows that apo and inhibitor-bound states produce opposite ruggedness directions vs the IDE M16 family.
σ range 0.08 → 1.70 · same family in all 40 runs · IDE vs NEP apo direction differs · locked ✓A benchmark suite containing only successes is scientifically weaker.
Prometheus locks both decision-grade successes and non-decision-grade constraint cases — explicitly, with canonical wording and public dataset entries. These locked diagnostic failures — ENG-001 through ENG-006 — document where the minimal-fragment model fails and why: Fe³⁺ and Mn²⁺ fragments collapse into the wrong spin sector with energy gaps so large that no penalty can recover the target state.
By retaining these failures, the benchmark demonstrates that ELSD promotion requires more than low final energy or nominal convergence. A platform that cannot say when a model is wrong is not a diagnostic — it is a marketing layer over successful runs.
Prometheus gains credibility not only by locking clean benchmark systems, but by refusing to lock fragment models that are physically wrong even when they converge numerically. ENG-001 through ENG-006 are not failures — they are proof that the filter works.
Rigid Stability confirmed across six coordination environments and four distinct inhibitor warhead classes. Metabolic campaign adds IDE (M16 Zn metallopeptidase, σ=0.32) and NEP (M13, σ=0.15) with three-state catalytic perturbation series — first S-donor thiolate result (σ=0.10, tightest Zn metallopeptidase result in the dataset). CA, MMP, and Zn metallopeptidase targets are rankable across all tested chemotypes.
VAP-1/AOC3 fully mapped across 8 states. N-donor inhibitors enter Rigid Stability; O-donor amplifies ruggedness. ADC chelator ladder locked: DOTA σ=0.45 outranks NOTA σ=0.51 — matching the known experimental hierarchy.
Fe²⁺ low-spin locked as Rigid Stability. Fe³⁺ and Mn²⁺ minimal fragments both fail with ~5–20 Ha spin-sector gaps — same failure class, same diagnosis. Benchmark-provisional pending QM/MM embedding.
ADC Chelator Electronic Stability Ladder
In Antibody-Drug Conjugates and radiopharmaceuticals, the chelator is the coordination cage that must hold the metal payload in a well-defined electronic state throughout delivery. Prometheus provides the first sector-audited electronic landscape benchmark for Cu²⁺ chelator coordination models — a new pre-experimental triage signal that sits upstream of costly animal stability testing.
Cu²⁺ DOTA — Industry Benchmark
DOTA is the gold-standard chelator for Cu-64 PET imaging agents and Cu-based ADC payloads. The square-planar 4N macrocyclic coordination state is electronically rigid and decision-grade. This is the audited baseline — any new chelator design must beat this number to credibly claim improved electronic stability.
Cu²⁺ NOTA — Smaller Macrocycle
NOTA is a widely-used but smaller, less preorganized chelator. It lands in Rigid Stability — decision-grade — but measurably broader than DOTA (12% higher σ). Prometheus independently reproduced the known experimental stability hierarchy between these two chelators using quantum geometry alone.
The key finding: Prometheus ranked DOTA above NOTA in electronic stability — matching the known experimental hierarchy — without any prior experimental input. This demonstrates that the ELSD framework captures a real physical signal about chelator coordination quality, not just numerical noise.
In practical terms, this means developers can rank candidate chelator coordination geometries earlier in the design process, identify electronically weaker designs before committing to synthesis, and focus experimental effort on architectures that are both chemically and electronically tractable.
Chemotype-Dependent Electronic Regime Shift
Eight coordination states of the VAP-1/AOC3 Cu²⁺ active site — four catalytic variants and four inhibitor coordination classes — all sector-audited and locked. The N-donor boundary is now mapped with three chemotypes on each side of the regime transition.
N-donor coordination at the VAP-1/AOC3 Cu²⁺ site is a robust electronic stabilization mechanism — confirmed across triazole, hydrazine, and imidazole: three chemically distinct classes with different ring size, heteroatom count, and pharmacophore profile. This is not an imidazole-specific finding. It is a donor-type finding. The O-donor semicarbazide σ=1.7903 sits ten times above the triazole σ=0.1747 — same target, same metal center, opposite electronic consequences. That boundary is mapped from both sides and is now defensible with six independent data points.
IDE + NEP Catalytic State Maps
Two metabolic Zn²⁺ metallopeptidase families with relevance to insulin, GLP-1, BNP, and amyloid-beta processing. Multi-state catalytic perturbation series locked for both — including warhead rankings and apo states. The S-donor thiolate at NEP (thiorphan class) is the tightest Zn metallopeptidase result in the dataset.
The warhead metal-specificity finding: At IDE (Zn²⁺, M16), the O-donor bidentate hydroxamate (σ=0.3988) outperforms the N-donor imidazole inhibitor (σ=0.4968). At NEP (Zn²⁺, M13), the S-donor thiolate (thiorphan, σ=0.0969) is the tightest result in the entire dataset. Metal center identity and family architecture both determine optimal warhead class — a distinction invisible to docking scores.
The apo-direction finding: IDE apo (σ=0.1473) is tighter than IDE resting (σ=0.3208). NEP apo (σ=0.2841) is broader than NEP resting (σ=0.1465). Same perturbation — water removal — produces opposite electronic consequences in the M13 vs M16 families. Prometheus detects this family-level difference.
Identify stable, decision-grade biological benchmarks
Fe, Zn, and Cu active sites — closed-shell and open-shell — classified with full sector audit across multiple seeds. Results survive scaffold enrichment and across four distinct warhead classes.
Fe σ=0.34 · Zn σ=0.09–0.51 · Cu⁺ σ=0.76 · all locked ✓Rank chelator and inhibitor designs by electronic tractability
DOTA outranks NOTA electronically. Triazole outranks semicarbazide ten-fold. S-donor thiolate at NEP produces the tightest Zn metallopeptidase result in the dataset. Prometheus independently reproduced known experimental rankings — without any prior experimental input.
DOTA σ=0.45 < NOTA σ=0.51 · Triazole σ=0.17 vs Semicarbazide σ=1.79 · NEP thiolate σ=0.10Characterize the electronic consequence of coordination mode — before synthesis
Two inhibitor coordination classes at the same Cu-centered VAP-1/AOC3 target produce a 10× difference in landscape ruggedness. IDE and NEP respond differently to the same perturbation (water removal). These distinctions are invisible to docking scores.
Same target · same metal · 10× difference · IDE vs NEP apo direction opposite · locked ✓The model validity
problem in drug discovery
Drug discovery teams already have ways to rank compounds. What they often lack is a way to verify that the metal-centered target model itself is trustworthy enough to rank against. A model that looks converged can still sit in the wrong electronic basin — producing rankings and SAR conclusions that are internally consistent but built on the wrong mechanistic foundation.
Prometheus does not just score a candidate pose. It characterizes the electronic consequence of a modeled coordination state — and whether that state is reproducible, tractable, and decision-grade. That is a different and upstream question from what docking answers.
Three-Part Cu Ruggedness Decomposition
Electronic landscape ruggedness in Cu-centered active sites is governed by three independently quantifiable contributions. Each demonstrated by controlled comparison on locked benchmark systems.
Open-shell d⁹ character is a primary ruggedness driver. Only oxidation state changes.
T-shaped geometry constrains the electronic manifold ~3× more tightly than pseudo-tetrahedral SOD.
Closed-shell d¹⁰ Zn²⁺ defines the Rigid Stability endpoint.
The Cu_AOC3_minimal fragment (SSAO/VAP-1, PDB 4BTY) is locked as a Coherent Open-Shell benchmark with 5/5 sector-clean seeds. AOC3's T-shaped Cu²⁺ geometry produces a materially tighter electronic landscape than the SOD pseudo-tetrahedral scaffold despite sharing the same d⁹ class and comparable 3N+1O donor composition. Foundation for the full 8-state VAP-1 electronic map.
Target Classification Matrix
Each card answers the same question before committing chemistry budget: is this target model electronically stable enough to rank against? Every σ value is derived from sector-audited, multi-seed ensembles on validated fragment models.
The metabolic warhead finding. At IDE Zn²⁺ (M16), O-donor bidentate hydroxamate (σ=0.3988) outperforms N-donor imidazole (σ=0.4968). At NEP Zn²⁺ (M13), the S-donor thiolate (σ=0.0969) is the tightest Zn metallopeptidase result in the dataset. Metal identity and family determine optimal warhead class — the N-donor rule established at Cu²⁺ VAP-1 does not transfer to Zn²⁺ targets.
The ADC chelator finding. Prometheus independently reproduced the known experimental stability hierarchy between DOTA and NOTA — DOTA (σ=0.45) outranking NOTA (σ=0.51) — using quantum geometry alone with no prior experimental input. This demonstrates that the ELSD framework captures a real physical signal about chelator coordination quality.
The N-donor rule at VAP-1. Three N-donor inhibitor classes — triazole, hydrazine, imidazole — all land in Rigid Stability (σ=0.17–0.32). The O-donor semicarbazide sits at σ=1.79. Same target, same metal center. The donor-type boundary is now mapped from both sides.
The coordination geometry finding. Cu²⁺ AOC3 T-shaped (σ=0.56) vs Cu²⁺ SOD pseudo-tetrahedral (σ=1.70) — same d⁹ class, ~3× difference in ruggedness. Local geometry is a major, separable contributor independent of oxidation state.
Most tools optimize on top of models they should not trust.
Prometheus works one layer earlier.
Drug programs around metalloproteins often fail not because the chemistry was wrong, but because the electronic model underpinning the target was never validated. Teams spend weeks optimizing compounds, building SAR narratives, commissioning assays — against a metal-center model that was unstable, multi-basin, or too truncated to support the conclusions drawn from it. Prometheus classifies that model before the chemistry budget is committed.
The result is not a faster drug. It is fewer false mechanistic leads, fewer wasted synthesis cycles, and faster prioritization of compounds and target-state models that are genuinely decision-grade.
The investable wedge is not "we simulate everything faster." It is: we prevent expensive programs from optimizing against the wrong electronic story. Prometheus works upstream — on whether the target-state model itself is decision-grade. That removes a category of failure that no existing tool addresses.
Prometheus fits between target model construction and ligand screening. It answers: is this a valid model for decision-making? Will small perturbations preserve the regime or flip the electronics? Are we screening against something coherent, or against an artifact? That question is extremely legible to a team that has lost months to a false mechanistic lead.
🛡 Three Expensive Failure Modes Prometheus Reduces
These are not edge cases. They are the reason metalloenzyme programs routinely run longer and cost more than anticipated.
False mechanistic commitment
Teams spend months on a story built on the wrong electronic state. The model converged — it just converged to the wrong basin. SAR, assay design, and compound prioritization all drift from the start.
Prometheus validates whether the modeled state is reproducible and sector-coherent before chemistry begins.Screening against artifacts
Thousands of compounds ranked against a metal-center model that is truncated, unstable, or multi-basin. The hits look real. The scaffold was the problem.
Prometheus classifies the model before large-scale screening — catching scaffold artifacts and fragment inadequacy early.Late discovery of target instability
Programs find too late that a protonation change, redox shift, or ligand substitution changes the active-site electronics so much that earlier SAR no longer transfers.
Prometheus maps when perturbations preserve regime versus induce a new electronic family — before medicinal chemistry is designed.📂 Four Categories That Matter Commercially
Each category has a different commercial implication. Prometheus has demonstrated the ability to distinguish all four — across Fe, Zn, and Cu systems validated with multiple independent seeds.
Decision-grade target models
Sector-clean, reproducible, single-family. Classification survives scaffold enrichment, warhead substitution, and inhibitor class variation. Includes three N-donor VAP-1 inhibitors, DOTA ADC chelator, and full metabolic Zn metallopeptidase catalytic state maps.
Executive translation: Prometheus identifies target models and chelator designs robust enough to support downstream screening with confidence.
Harder systems that are still trustworthy
Broader ensemble than closed-shell controls, but sector-clean and single-family. Open-shell does not mean unusable — the bar is higher and Prometheus can still clear it. This category spans a wide tractability range within the same regime.
Executive translation: Prometheus classifies realistic open-shell targets and tells you when they are trustworthy enough to guide work.
Systems that look simple but are genuinely ambiguous
Sector-clean but multiple competing orbital families. Not a bad run — an intrinsically ambiguous scaffold. Without detection, this becomes the foundation for a misleading screening campaign.
Executive translation: Prometheus flags targets where the electronic landscape is genuinely multi-basin before teams over-interpret a single optimized state.
Models too truncated to support any conclusion
Sector escape or spin escalation despite strong enforcement. The fragment is physically wrong — any result would be an artifact. These locked failures are proof that the filter works: Prometheus refuses to certify models that are numerically converged but physically incorrect.
Executive translation: Prometheus tells you when a model should not be used for decision-making at all. That is the most commercially valuable thing a quality-control tool can say.
⚡ The Challenge Standard Methods Cannot Fully Resolve
The active sites of P450s, SOD, carbonic anhydrase, SSAO/VAP-1, ADC chelators, Zn metallopeptidases (IDE, NEP), and many other drug targets contain transition metals with electronic regimes that are difficult to classify reliably using conventional workflows alone.
The gap in current workflows
Standard simulation finds the lowest energy state — it does not characterize the full set of accessible electronic configurations, or detect whether the model is sitting in the correct basin. For open-shell metallocenters and coordination cages, that gap matters.
What Prometheus adds
A reproducible, sector-audited ensemble sweep that classifies the electronic landscape — stable, coherent open-shell, multi-basin, or model-pathologic — before any synthesis decision is made. Not a replacement for existing workflows. A quality-control layer upstream of them.
📐 The Prometheus Classification System
Prometheus does not merely find energy minima. It sorts metallocenters into mechanistically distinct regimes — and identifies when a fragment model is too truncated to support a trustworthy classification.
Rigid Stability
Low ensemble dispersion, single-family basin, sector-clean across all seeds. The model is decision-grade and reproducible.
typically sub-kcal/mol σCoherent Open-Shell
Broader ensemble than closed-shell controls, but sector-clean and single-family. Characteristic of realistic open-shell transition-metal targets.
broader but still coherentMulti-Basin
Sector-clean but multiple competing orbital families. Scaffold geometry is the causal variable. Indicates a genuinely ambiguous electronic environment.
large dispersion / multiple familiesModel Pathology
Sector escape or spin escalation despite strong enforcement. Fragment model is physically wrong. No result from this model should guide decisions.
engineering constraint🧲 Zn Warhead Rankability Ladder
Four chemically distinct Zn-binding warhead classes tested on the same CA2-like scaffold. All four land in Rigid Stability. Scaffold geometry, not warhead electronics, sets the ruggedness floor.
MMP / CA relevant
weaker Zn binder
HDAC / CA scaffold
CA IX/XII inhibitor class
All four warhead classes: Rigid Stability. The CA-like Zn active site is rankable regardless of inhibitor chemotype. This makes Zn-centered CA and MMP targets strong candidates for high-confidence lead optimization independent of zinc-binding warhead class.
36 locked canonical benchmark and diagnostic results
| System | Metal / Config | σ (kcal/mol) | Seeds | Classification | Status |
|---|---|---|---|---|---|
| FePorphyrin_FeII lsP450 / heme model | Fe²⁺ d⁶ low-spin | 0.3375 | 5 | Rigid Stability | ● LOCKED |
| Zn_CA2 minimalCarbonic anhydrase | Zn²⁺ d¹⁰ tetrahedral | 0.0933 | 9 | Rigid Stability | ● LOCKED |
| Zn_CA2 imidazoleFull His coordination shell | Zn²⁺ d¹⁰ + 3 imidazole | 0.4367 | 5 | Rigid Stability | ● LOCKED |
| Zn_CA2 sulfonamideCA IX/XII inhibitor class | Zn²⁺ d¹⁰ + sulfonamide | 0.5058 | 5 | Rigid Stability | ● LOCKED |
| Zn_CA2 hydroxamateHDAC / CA scaffold class | Zn²⁺ d¹⁰ + hydroxamate | 0.4672 | 5 | Rigid Stability | ● LOCKED |
| Zn_CA2 carboxylateWeaker Zn-binding chemotype | Zn²⁺ d¹⁰ + formate | 0.3490 | 5 | Rigid Stability | ● LOCKED |
| Zn_CA2 phosphonateMMP / CA scaffold class | Zn²⁺ d¹⁰ + phosphonate | 0.3307 | 5 | Rigid Stability | ● LOCKED |
| Cu_SOD_minimalSuperoxide dismutase baseline | Cu²⁺ d⁹ doublet | 1.7035 | 5 | Coherent Open-Shell | ● LOCKED |
| Cu_SOD_minimal_waterFull scaffold + axial H₂O | Cu²⁺ d⁹ + H₂O | 0.9414 | 5 | Coherent Open-Shell | ● LOCKED |
| Cu_SOD_minimal_CuId¹⁰ oxidation-state control | Cu⁺ d¹⁰ — same scaffold | 0.7592 | 5 | Intermediate / Controlled | ● LOCKED |
| Cu_SOD_2imidazoleApo — coord. number 4→3 | Cu²⁺ d⁹ 2N+1O | 0.0803 | 5 | Rigid Stability | ● LOCKED |
| Cu_SOD_2imidazole_waterReduced scaffold + axial H₂O | Cu²⁺ d⁹ + H₂O axial | 0.7419 | 5 | Coherent Open-Shell | ● LOCKED |
| Cu_SOD_protonatedDonor neutralized | Cu²⁺ d⁹ formic acid | 0.6402 | 5 | Coherent Open-Shell | ● LOCKED |
| Cu_SOD_acetateDonor strengthened | Cu²⁺ d⁹ acetate | 0.9017 | 5 | Coherent Open-Shell | ● LOCKED |
| Cu_AOC3_minimalSSAO/VAP-1 · resting state | Cu²⁺ d⁹ T-shaped | 0.5610 | 5 | Coherent Open-Shell | ● LOCKED |
| Cu_AOC3_minimal_apoVAP-1 · true apo state | Cu²⁺ d⁹ T-shaped, no axial | 0.4854 | 5 | Rigid Stability | ● LOCKED |
| Cu_AOC3_minimal_protonatedVAP-1 · donor weakened | Cu²⁺ d⁹ protonated His | 0.5536 | 5 | Coherent Open-Shell | ● LOCKED |
| Cu_AOC3_minimal_hydroxideVAP-1 · catalytic OH⁻ | Cu²⁺ d⁹ + OH⁻ axial | 0.5965 | 5 | Coherent Open-Shell | ● LOCKED |
| Cu_AOC3_imidazoleVAP-1 · N-donor inhibitor | Cu²⁺ d⁹ + imidazole axial | 0.3235 | 5 | Rigid Stability | ● LOCKED |
| Cu_AOC3_semicarbazideVAP-1 · O-donor inhibitor | Cu²⁺ d⁹ + semicarbazide | 1.7903 | 5 | Coherent Open-Shell | ● LOCKED |
| Cu_AOC3_triazoleVAP-1 · triazole N-donor | Cu²⁺ d⁹ + triazole axial | 0.1747 | 5 | Rigid Stability | ● LOCKED |
| Cu_AOC3_hydrazineVAP-1 · hydrazine N-donor | Cu²⁺ d⁹ + hydrazine axial | 0.2003 | 5 | Rigid Stability | ● LOCKED |
| Cu_DOTA_minimalADC chelator · DOTA 4N macrocycle | Cu²⁺ d⁹ square-planar 4N | 0.4534 | 5 | Rigid Stability | ● LOCKED |
| Cu_NOTA_minimalADC chelator · NOTA 3N macrocycle | Cu²⁺ d⁹ trigonal 3N | 0.5095 | 5 | Rigid Stability | ● LOCKED |
| Metabolic · IDE (Insulin-Degrading Enzyme) · M16 Zn Peptidase · His/His/Glu/water | |||||
| IDE_minimalGLP-1 degradation gateway · resting | Zn²⁺ d¹⁰ 2N+1O+1O | 0.3208 | 5 | Rigid Stability | ● LOCKED |
| IDE_minimal_aqua3-coordinate apo state | Zn²⁺ d¹⁰ 2N+1O | 0.1473 | 5 | Rigid Stability | ● LOCKED |
| IDE_minimal_hydroxamateO-donor bidentate warhead | Zn²⁺ d¹⁰ + hydroxamate | 0.3988 | 5 | Rigid Stability | ● LOCKED |
| IDE_minimal_hydroxideCatalytic OH⁻ attack state | Zn²⁺ d¹⁰ + OH⁻ | 0.4310 | 5 | Rigid Stability | ● LOCKED |
| IDE_minimal_inhibitorN-donor imidazole warhead | Zn²⁺ d¹⁰ + imidazole | 0.4968 | 5 | Rigid Stability | ● LOCKED |
| Metabolic · NEP (Neprilysin) · M13 Zn Peptidase · His/His/Glu/water | |||||
| NEP_minimal_thiorphanS-donor thiolate · tightest Zn result | Zn²⁺ d¹⁰ + thiolate S | 0.0969 | 5 | Rigid Stability | ● LOCKED |
| NEP_minimalWater-bound resting state | Zn²⁺ d¹⁰ 2N+1O+1O | 0.1465 | 5 | Rigid Stability | ● LOCKED |
| NEP_minimal_apo3-coordinate apo state | Zn²⁺ d¹⁰ 2N+1O | 0.2841 | 5 | Rigid Stability | ● LOCKED |
| Zn_squareplanarScaffold geometry control | Zn²⁺ d¹⁰ D4h artificial | 43.29 | 5 | Multi-basin (scaffold) | ◆ DIAGNOSTIC |
20-qubit penalized VQE · LANL2DZ basis · GPU-accelerated L40S · Full sector audit per run · Zenodo v1: 10.5281/zenodo.19142883 · Zenodo v2: 10.5281/zenodo.19163956 · Zenodo v3: 10.5281/zenodo.19212485 · JCTC under consideration ct-2026-00566z
🧪 Cu²⁺ SOD Perturbation Series
A five-experiment controlled perturbation panel on a locked Cu²⁺ SOD active-site model. One variable changed per experiment. Same dominant electronic family in all 40 runs. σ varied from 0.08 to 1.70 kcal/mol — a 21× range driven entirely by first-shell chemical changes.
Water binding on the reduced scaffold increased ruggedness 9.2× (σ: 0.08 → 0.74). Water binding on the full baseline scaffold tightened it 45% (σ: 1.70 → 0.94). Both events preserve the dominant electronic family. A discovery program that treats them as interchangeable is operating with the wrong model.
🔬 How It Works
A quality-control layer, not an optimizer. Prometheus answers a different question from standard simulation: is this electronic model trustworthy enough to guide chemistry?
Fragment model construction
First-shell metallocenter geometry built from crystallographic references. Each atom verified; engine backed up before any run.
Penalized VQE multi-seed sweep
Independent VQE optimizations across 5 seeds with N-sector and Sz-sector enforcement. 20-qubit operating point validated across all targets.
Sector audit per seed
Every run: full energy decomposition, ⟨N⟩, ⟨Sz⟩, dom_p, dominant bitstring. Penalty contributions explicitly separated from physical energy.
Landscape classification
σ across seeds, basin count, dominant determinant consistency. Result locked only when all metrics are internally consistent across the full ensemble.
Realism-survival testing
Classification re-run with enriched scaffold and across multiple warhead or inhibitor classes. For Zn: 4 chemotypes all Rigid Stability. For AOC3: 8 coordination states mapped. For ADC: DOTA/NOTA ladder confirmed. For metabolic Zn peptidases: multi-state benchmark series locked for IDE and NEP.
What makes a result trustworthy?
Every locked Prometheus result meets all five criteria simultaneously — not a subset.
"Prometheus gains credibility not only by locking clean benchmark systems, but by refusing to lock fragment models that are physically wrong even when they converge numerically."
Research Roadmap
36 locked canonical results across three therapeutic areas. Zenodo v1, v2, and v3 published. JCTC manuscript under consideration (ct-2026-00566z). VAP-1 full 8-state map complete. ADC chelator ladder established. Metabolic Zn peptidase campaign established: IDE (M16) and NEP (M13) multi-state benchmark series and warhead contrasts locked. Prometheus is available for partner target classification engagements.
✓ Complete
Biological Benchmark Ladder
36 locked results across Fe, Zn, and Cu. Zn: 4 warhead classes Rigid Stability + Metabolic Zn peptidase families (IDE M16, NEP M13) with multi-state benchmark series. Cu: three-part decomposition + 8-state VAP-1 map + ADC chelator ladder (DOTA/NOTA). Fe and Mn: 6 engineering constraints documenting minimal-fragment limits.
✓ Complete · Under Review
Mechanistic Validation & Publication
Three-part Cu decomposition complete. VAP-1 N-donor boundary mapped from both sides. ADC chelator hierarchy independently reproduced. Metabolic warhead specificity: O-donor bidentate optimal for IDE Zn²⁺; S-donor thiolate produces tightest Zn metallopeptidase result for NEP. Manuscript under consideration at JCTC (ct-2026-00566z). Zenodo v1, v2, and v3 published.
◆ Now Available
Partner Target Classification
Apply the validated Prometheus workflow to partner-nominated metalloenzyme targets or chelator coordination models. Deliver a sector-audited electronic landscape report — decision-grade, coherent open-shell, multi-basin, or model-pathologic — before screening begins.
Stop optimizing against models you haven't validated.
Prometheus is available for a small number of partner engagements in 2026. If your program involves a metal-centered target or ADC chelator and you need to know whether the working electronic model is stable, coherent, multi-basin, or too truncated to trust — before you commit to large-scale screening or medicinal chemistry — we should talk.
Typical engagement: sector-audited electronic landscape report for your target system, typically delivered in days depending on target complexity.