Prometheus Platform  ·  Biological Domain  ·  March 2026

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.

To our knowledge, the first sector-audited, multi-start electronic-landscape characterization of the VAP-1/AOC3 Cu²⁺ active site across six catalytic and inhibitor-bound states — and the first audited electronic benchmark for DOTA-class ADC chelators and Zn metallopeptidase catalytic cycle states. Same target, same metal center, opposite electronic consequences depending on chemotype.
What we deliver

Sector-audited electronic landscape classification for metalloenzyme active sites — in days, across Fe, Zn, and Cu systems, with reproducible multi-seed validation.

Validated benchmark families

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)

36 Locked Canonical Results Fe · Zn · Cu Systems 3 Therapeutic Areas Three-Part Mechanistic Decomposition VAP-1 Full Map · 8 States ADC Chelator Ladder Metabolic · IDE + NEP Series
In plain English — no chemistry degree required

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.

🎯 Oncology · ADC Cancer Treatments

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.

🧠 Neuroinflammation · VAP-1 / SSAO

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.

The analogy that explains everything

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.

Executive Decision Tool

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

The three questions that matter before screening begins

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.

Is this target model electronically decision-grade?

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 ✓
Does my inhibitor or chelator coordination mode stabilize or amplify the electronic landscape?

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=0.10
Will a perturbation preserve the electronic regime or flip it?

A 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 ✓
Platform Credibility

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.

Zinc · d¹⁰

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.

Copper · VAP-1 Map + ADC Chelator Ladder

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.

Iron & Manganese · Provisional

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 Oncology · First Chelator Electronic Ladder · March 2026

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.

✓ Gold Standard · 12-Membered Ring · 4N Donors

Cu²⁺ DOTA — Industry Benchmark

0.4534
Rigid Stability · 5/5 seeds clean · dom_p 0.9990–0.9999

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.

◆ Contrast · 9-Membered Ring · 3N Donors

Cu²⁺ NOTA — Smaller Macrocycle

0.5095
Rigid Stability · 5/5 seeds clean · dom_p 0.9990–1.0000

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.

VAP-1 / SSAO-AOC3 · Complete Electronic Map · 8 States · March 2026

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 · Triazole
1,2,4-Triazole
0.1747
Rigid Stability
N-Donor · Hydrazine
Hydrazine
0.2003
Rigid Stability
N-Donor · Imidazole
Imidazole
0.3235
Rigid Stability
True Apo
No axial ligand
0.4854
Rigid Stability
Donor Weakened
Protonated
0.5536
Coherent Open-Shell
Water-Bound
Resting state
0.5610
Coherent Open-Shell
Catalytic State
Hydroxide OH⁻
0.5965
Coherent Open-Shell
O-Donor Inhibitor
Semicarbazide
1.7903
Elevated Ruggedness
Same target. Same metal center. Opposite electronic consequences.

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.

Metabolic Disease · Zn Metallopeptidase Campaign · March 2026

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.

IDE (Insulin-Degrading Enzyme) · M16 Zn Peptidase · His/His/Glu/water · σ range 0.15–0.50
3-coord Apo
Apo (water removed)
0.1473
Rigid Stability
Resting State
Water-bound
0.3208
Rigid Stability
O-donor Warhead
Hydroxamate
0.3988
Rigid Stability
Catalytic OH⁻
Hydroxide
0.4310
Rigid Stability
N-donor Warhead
Imidazole inhibitor
0.4968
Rigid Stability
NEP (Neprilysin) · M13 Zn Peptidase · His/His/Glu/water · σ range 0.10–0.28
S-donor · Tightest Zn result
Thiorphan thiolate
0.0969
Rigid Stability · new donor class
Resting State
Water-bound
0.1465
Rigid Stability
3-coord Apo
Apo (water removed)
0.2841
Rigid Stability · apo broader than resting

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.

Three things Prometheus demonstrates that competitors blur together
1

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 ✓
2

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=0.10
3

Characterize 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 ✓
Why This Matters

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.

Axis 1 · Oxidation State
Cu²⁺ SOD d⁹ vs Cu⁺ SOD d¹⁰
↓ same scaffold ↓
σ: 1.7035 → 0.7592 kcal/mol

Open-shell d⁹ character is a primary ruggedness driver. Only oxidation state changes.

Axis 2 · Coordination Geometry
Cu²⁺ SOD pseudo-tet vs Cu²⁺ AOC3 T-shaped
↓ same d⁹ class ↓
σ: 1.7035 → 0.5610 kcal/mol

T-shaped geometry constrains the electronic manifold ~3× more tightly than pseudo-tetrahedral SOD.

Axis 3 · Metal Identity
Cu²⁺/Cu⁺ open-shell vs Zn²⁺ d¹⁰ closed-shell
↓ metal identity ↓
σ: 0.56–1.70 → 0.09–0.51 kcal/mol

Closed-shell d¹⁰ Zn²⁺ defines the Rigid Stability endpoint.

First Metabolic-Disease Benchmark · Locked March 22, 2026
Cu²⁺ AOC3 / SSAO-VAP-1 — Locked Coherent Open-Shell Baseline

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.

σ baseline
0.5610
Seeds
5/5 clean
dom_p
0.9993
vs SOD
3× tighter
States mapped
8 total
Risk Assessment Tool · March 2026

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.

d¹⁰ · Zn²⁺
Zn CA2 · 4 warhead classes
0.33–0.51
Rigid Stability
Safe to Rank
Sulfonamide · hydroxamate · phosphonate · carboxylate — all rigid
d⁶ · Fe²⁺ ls
Fe Porphyrin · heme model
0.34
Rigid Stability
Safe to Rank
Low-spin Fe²⁺ locked; Fe branch otherwise provisional
d⁹ · Cu²⁺ · N-donor inhibitors
VAP-1 + Triazole / Hydrazine / Imidazole
0.17–0.32
Rigid Stability
Safe to Rank
All three N-donor classes confirm the same regime — robust finding
d⁹ · Cu²⁺ · DOTA 4N macrocycle
ADC Chelator · Gold Standard
0.4534
Rigid Stability
Safe to Rank
First ADC chelator benchmark — electronic gold standard confirmed
d¹⁰ · Zn²⁺ · M16 Zn peptidase
IDE · GLP-1 degradation gateway
0.15–0.50
Rigid Stability · 5-state map
Safe to Rank
Full catalytic cycle locked · O-donor bidentate (hydroxamate) outperforms N-donor at IDE Zn²⁺
d¹⁰ · Zn²⁺ · M13 Zn peptidase
NEP · GLP-1 / BNP degradation
0.10–0.28
Rigid Stability · 3-state map
Safe to Rank
S-donor thiolate (σ=0.0969) tightest Zn metallopeptidase result in the dataset · first S-donor warhead class
d⁹ · Cu²⁺ · NOTA 3N macrocycle
ADC Chelator · Contrast
0.5095
Rigid Stability
Rank with Care
Decision-grade but broader than DOTA — matches known experimental hierarchy
d⁹ · Cu²⁺ · T-shaped
Cu AOC3 / SSAO-VAP-1 baseline
0.56
Coherent Open-Shell
Rank with Care
3× tighter than SOD — T-shaped geometry constrains accessible manifold
d¹⁰ · Cu⁺
Cu SOD · reduced state
0.76
Intermediate / Controlled
Safe to Rank
Above Zn floor; geometry contributes residually
d⁹ · Cu²⁺ · O-donor inhibitor
AOC3 + Semicarbazide-class
1.7903
Coherent Open-Shell
QC Before Ranking
Classical SSAO inhibitor class amplifies ruggedness to SOD-baseline levels

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.

The Business Case

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.


For investors

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.

For R&D leadership

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.

Category 1 · Stable

Decision-grade target models

FePorphyrin FeII ls · Zn CA2 (4 warhead classes) · VAP-1 N-donor inhibitors · Cu DOTA chelator · IDE (M16) · NEP (M13)

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.

Category 2 · Coherent Open-Shell

Harder systems that are still trustworthy

Cu_SOD_minimal σ=1.70 · Cu_AOC3 baseline σ=0.56 · Cu NOTA σ=0.51 · Semicarbazide σ=1.79

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.

Category 3 · Multi-Basin

Systems that look simple but are genuinely ambiguous

Zn_squareplanar · σ = 43.29 kcal/mol · scaffold artifact

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.

Category 4 · Pathologic

Models too truncated to support any conclusion

ENG-001 through ENG-006 · Fe and Mn branch constraints + insulin Zn hexamer

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.

01

Rigid Stability

Low ensemble dispersion, single-family basin, sector-clean across all seeds. The model is decision-grade and reproducible.

typically sub-kcal/mol σ
02

Coherent 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 coherent
03

Multi-Basin

Sector-clean but multiple competing orbital families. Scaffold geometry is the causal variable. Indicates a genuinely ambiguous electronic environment.

large dispersion / multiple families
04

Model 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.

Phosphonate
Methylphosphonate
MMP / CA relevant
0.3307
kcal/mol · 5 seeds · locked
Carboxylate
Formate
weaker Zn binder
0.3490
kcal/mol · 5 seeds · locked
Hydroxamate
Acetohydroxamate
HDAC / CA scaffold
0.4672
kcal/mol · 5 seeds · locked
Sulfonamide
Sulfonamide-NH
CA IX/XII inhibitor class
0.5058
kcal/mol · 5 seeds · locked
Validated Results · March 2026

36 locked canonical benchmark and diagnostic results

Every locked result is reproducible across multiple seeds with full sector and energy decomposition. Diagnostic controls and engineering constraints included explicitly — failures are as important as successes for platform credibility.
SystemMetal / Configσ (kcal/mol)SeedsClassificationStatus
FePorphyrin_FeII lsP450 / heme modelFe²⁺ d⁶ low-spin0.33755Rigid Stability● LOCKED
Zn_CA2 minimalCarbonic anhydraseZn²⁺ d¹⁰ tetrahedral0.09339Rigid Stability● LOCKED
Zn_CA2 imidazoleFull His coordination shellZn²⁺ d¹⁰ + 3 imidazole0.43675Rigid Stability● LOCKED
Zn_CA2 sulfonamideCA IX/XII inhibitor classZn²⁺ d¹⁰ + sulfonamide0.50585Rigid Stability● LOCKED
Zn_CA2 hydroxamateHDAC / CA scaffold classZn²⁺ d¹⁰ + hydroxamate0.46725Rigid Stability● LOCKED
Zn_CA2 carboxylateWeaker Zn-binding chemotypeZn²⁺ d¹⁰ + formate0.34905Rigid Stability● LOCKED
Zn_CA2 phosphonateMMP / CA scaffold classZn²⁺ d¹⁰ + phosphonate0.33075Rigid Stability● LOCKED
Cu_SOD_minimalSuperoxide dismutase baselineCu²⁺ d⁹ doublet1.70355Coherent Open-Shell● LOCKED
Cu_SOD_minimal_waterFull scaffold + axial H₂OCu²⁺ d⁹ + H₂O0.94145Coherent Open-Shell● LOCKED
Cu_SOD_minimal_CuId¹⁰ oxidation-state controlCu⁺ d¹⁰ — same scaffold0.75925Intermediate / Controlled● LOCKED
Cu_SOD_2imidazoleApo — coord. number 4→3Cu²⁺ d⁹ 2N+1O0.08035Rigid Stability● LOCKED
Cu_SOD_2imidazole_waterReduced scaffold + axial H₂OCu²⁺ d⁹ + H₂O axial0.74195Coherent Open-Shell● LOCKED
Cu_SOD_protonatedDonor neutralizedCu²⁺ d⁹ formic acid0.64025Coherent Open-Shell● LOCKED
Cu_SOD_acetateDonor strengthenedCu²⁺ d⁹ acetate0.90175Coherent Open-Shell● LOCKED
Cu_AOC3_minimalSSAO/VAP-1 · resting stateCu²⁺ d⁹ T-shaped0.56105Coherent Open-Shell● LOCKED
Cu_AOC3_minimal_apoVAP-1 · true apo stateCu²⁺ d⁹ T-shaped, no axial0.48545Rigid Stability● LOCKED
Cu_AOC3_minimal_protonatedVAP-1 · donor weakenedCu²⁺ d⁹ protonated His0.55365Coherent Open-Shell● LOCKED
Cu_AOC3_minimal_hydroxideVAP-1 · catalytic OH⁻Cu²⁺ d⁹ + OH⁻ axial0.59655Coherent Open-Shell● LOCKED
Cu_AOC3_imidazoleVAP-1 · N-donor inhibitorCu²⁺ d⁹ + imidazole axial0.32355Rigid Stability● LOCKED
Cu_AOC3_semicarbazideVAP-1 · O-donor inhibitorCu²⁺ d⁹ + semicarbazide1.79035Coherent Open-Shell● LOCKED
Cu_AOC3_triazoleVAP-1 · triazole N-donorCu²⁺ d⁹ + triazole axial0.17475Rigid Stability● LOCKED
Cu_AOC3_hydrazineVAP-1 · hydrazine N-donorCu²⁺ d⁹ + hydrazine axial0.20035Rigid Stability● LOCKED
Cu_DOTA_minimalADC chelator · DOTA 4N macrocycleCu²⁺ d⁹ square-planar 4N0.45345Rigid Stability● LOCKED
Cu_NOTA_minimalADC chelator · NOTA 3N macrocycleCu²⁺ d⁹ trigonal 3N0.50955Rigid Stability● LOCKED
Metabolic · IDE (Insulin-Degrading Enzyme) · M16 Zn Peptidase · His/His/Glu/water
IDE_minimalGLP-1 degradation gateway · restingZn²⁺ d¹⁰ 2N+1O+1O0.32085Rigid Stability● LOCKED
IDE_minimal_aqua3-coordinate apo stateZn²⁺ d¹⁰ 2N+1O0.14735Rigid Stability● LOCKED
IDE_minimal_hydroxamateO-donor bidentate warheadZn²⁺ d¹⁰ + hydroxamate0.39885Rigid Stability● LOCKED
IDE_minimal_hydroxideCatalytic OH⁻ attack stateZn²⁺ d¹⁰ + OH⁻0.43105Rigid Stability● LOCKED
IDE_minimal_inhibitorN-donor imidazole warheadZn²⁺ d¹⁰ + imidazole0.49685Rigid Stability● LOCKED
Metabolic · NEP (Neprilysin) · M13 Zn Peptidase · His/His/Glu/water
NEP_minimal_thiorphanS-donor thiolate · tightest Zn resultZn²⁺ d¹⁰ + thiolate S0.09695Rigid Stability● LOCKED
NEP_minimalWater-bound resting stateZn²⁺ d¹⁰ 2N+1O+1O0.14655Rigid Stability● LOCKED
NEP_minimal_apo3-coordinate apo stateZn²⁺ d¹⁰ 2N+1O0.28415Rigid Stability● LOCKED
Zn_squareplanarScaffold geometry controlZn²⁺ d¹⁰ D4h artificial43.295Multi-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.

Exp 3 · Apo
2-imidazole
0.08
Rigid Stability
Exp 1 · Donor −
Protonated
0.64
Tightened
Exp 2 · Donor +
Acetate
0.90
Tightened
Baseline
3N+1O formate
1.70
Coherent open-shell
Exp 4 · Bound
+ Water (reduced)
0.74
Coherent open-shell
The binding context finding
Same ligand. Opposite ruggedness direction. Different coordination context.

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?

01

Fragment model construction

First-shell metallocenter geometry built from crystallographic references. Each atom verified; engine backed up before any run.

02

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.

03

Sector audit per seed

Every run: full energy decomposition, ⟨N⟩, ⟨Sz⟩, dom_p, dominant bitstring. Penalty contributions explicitly separated from physical energy.

04

Landscape classification

σ across seeds, basin count, dominant determinant consistency. Result locked only when all metrics are internally consistent across the full ensemble.

05

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.

Sector clean (⟨N⟩, ⟨Sz⟩)±0.1 tolerance
Dominant determinant probabilitydom_p > 0.99
Same dominant familyall seeds
Penalty contribution auditeddecomposed per run
Realism-survival testedZn: 4 warheads · AOC3: 8 states · ADC: 2 chelators · metabolic: IDE + NEP multi-state series

"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.