QC Ware and IonQ Mapped Four Orbitals of a Cytochrome P450nor Active Site Onto Eight Qubits of IonQ Forte on 1 September 2026
Quentir Medicine Monitor
Evidence-based insights for quantum medicine. Published by Quentir Systems LLC · September 1, 2026.
How tightly a candidate binds its target is one important factor in whether a drug works, and metal centers are among the difficult cases to calculate: an iron atom sitting in the middle of an enzyme. Quantum chemists have spent decades building approximations around those centers, because the electrons there are correlated in ways that ordinary methods handle badly. On 1 September 2026, two companies said they had measured a four-orbital slice of one such site on a quantum computer and landed close to the classical answer.
QC Ware announced the result through a PR Newswire release issued on 1 September 2026. Its Promethium platform did the classical preparation, IonQ's Forte trapped-ion processor took the quantum measurement through Amazon Braket, and the target was the heme active site of cytochrome P450nor. The workflow calculated an electrostatic interaction energy within 0.5 kcal/mol of classical benchmarks, roughly four percent, which sits inside the one kcal/mol threshold conventionally called chemical accuracy. Amazon Web Services supported the run with cloud compute credits.
Practical takeaway. The quantum step here ran on eight qubits of a 36-qubit machine, measuring a four-orbital slice carved out of a 115-atom model that contains more than a thousand molecular orbitals. Everything before and after that slice happened on classical hardware. This is the second run of the same QC Ware workflow on a different machine, after an IBM Heron demonstration in August.
This is the IonQ half of a workflow this Monitor already read on IBM hardware in August
The release is a follow-up, and it says so obliquely. Sham's quotation refers to "our recent demonstration on other quantum hardware," and that earlier demonstration is one this Monitor examined at the time. On 7 August 2026 we published a reading of QC Ware's Promethium workflow running a nitric oxide reductase calculation on IBM's 156-qubit Heron processor, based on the company's PR Newswire release of that demonstration. The platform, the enzyme family and the hybrid division of labor are the same. What has changed is the processor underneath.
That makes the portability claim the actual news, and it also sets what today's release has to add to be worth reading twice. It adds four things the August account did not carry: the 115-atom size of the active-site model, the count of more than 1,000 molecular orbitals in it, the reduction to a four-orbital active space mapped onto eight qubits, and a numerical comparison against classical benchmarks expressed as 0.5 kcal/mol, roughly four percent. Those are real additions to the public record. They are also the numbers that let a reader size the quantum contribution for the first time, which the August release did not permit.
What QC Ware and IonQ announced on 1 September 2026, and what the numbers say
The release is specific about the pipeline, which makes it possible to read. Promethium built a 115-atom model of the P450nor active site containing over 1,000 molecular orbitals. It then identified the strongly correlated region of that model and reduced it to a four-orbital active space, which was mapped onto eight qubits. IonQ Forte measured those eight qubits in a single basis and returned the results, and Promethium computed the final interaction energies classically.
Two figures carry the claim. The first is the 0.5 kcal/mol agreement with classical benchmarks. The second is the statement that the hybrid workflow delivers more than double the accuracy of the standard classical mean-field method. Mean-field methods treat each electron as moving in the averaged field of all the others, and they are known to struggle exactly where this experiment was aimed, at an open-shell iron center with near-degenerate orbitals. Beating a mean-field baseline by a factor of two at such a site is a real result and a modest one at the same time, because mean-field is the weaker of the two comparators in play. The stronger one, whatever produced the classical benchmarks the 0.5 kcal/mol figure is measured against, is never identified in the release.
Kin-Joe Sham, co-founder and chief operating officer at QC Ware, framed the point of the exercise as hardware independence, saying that running the same workflow on trapped ions after an earlier demonstration on different quantum hardware shows the approach "is not tied to a single type of quantum hardware." Scott Millard, chief business officer at IonQ, put the commercial case in clinical terms, arguing that hybrid workflows can now "predict certain binding behavior accurately enough for discovery teams to confidently rank candidates and catch toxicity risks early." The qualifier "certain" is doing quiet work in that sentence, and the release does not say which behaviors fall inside it.
Which enzyme this actually was: P450nor is a fungal denitrification cytochrome
The choice of molecule deserves attention, because the release's own framing invites a misreading. It describes P450nor as "a nitric oxide reductase in the cytochrome P450 superfamily, the same superfamily whose monooxygenase members carry out most human drug metabolism." That sentence is accurate. It is also easy to compress, in a second-hand summary, into the false idea that a human drug-metabolizing enzyme was modeled.
It was not. As Hirofumi Shoun and colleagues set out in Philosophical Transactions of the Royal Society B in 2012, P450nor is a fungal CYP55 nitric oxide reductase, characterized in Fusarium oxysporum among other fungi. It reduces nitric oxide to nitrous oxide during anaerobic respiration, and it takes its electrons directly from reduced nicotinamide adenine dinucleotide, NADH, rather than through the reductase partner a human monooxygenase depends on. The release names no organism and no accession, so the species behind the modeled structure is not established by the announcement itself. Shoun and colleagues describe an enzyme that has been characterized in detail across its structure, its iron ligand configuration and its reaction mechanism, which is a good reason to pick it for a hardware demonstration and a poor reason to describe the run as modeling human drug metabolism.
Quantum pillar: computing. Technology readiness: TRL 3 of 9. This was a single run on hardware that exists: eight qubits measured a four-orbital slice of one model enzyme and the answer was checked against a classical calculation, which sits well below any test inside a real drug discovery program and far below anything a regulator would see.
How eight of IonQ Forte's 36 qubits carried the quantum step
IonQ publishes Forte's specifications: 36 qubits, all-to-all connectivity, a one-qubit gate error of 0.02 percent, a two-qubit gate error of 0.4 percent and a state preparation and measurement error of 0.5 percent. The release credits that all-to-all connectivity for letting the entangling gates run as designed without the routing overhead a limited-connectivity architecture would add, and on a trapped-ion machine that is a fair claim. The configuration is unusually compact for a hardware demonstration: eight qubits and a single measurement basis. What that compactness implies about error sources cannot be settled from the release, which gives no circuit, depth, compiler, shot count or mitigation detail.
The compression ratio is the number worth holding onto. A 115-atom model with more than a thousand molecular orbitals was reduced to four orbitals for the quantum machine to touch. Active-space selection of that kind is standard practice in computational chemistry and it is where most of the physics decisions get made. Choosing which four orbitals are strongly correlated is a classical judgment, made by classical software, and the quality of the final energy also depends materially on it, alongside what happens in the eight qubits afterward. Promethium automates that selection, which the release presents as a feature, and automation of a modeling choice is also a place where an error becomes invisible.
Why the benchmark comparison bounds the claim
An agreement with a classical benchmark is a validation result, and validation results run in one direction only. The classical number existed first, and the companies report agreement for this known case. What follows from that is narrower than it looks: the pipeline returned the expected answer once, on a system where the expected answer was available. Whether it returns a trustworthy number where no classical answer exists is a separate question, and it is the only question a discovery team would pay to have settled.
This is the recurring shape of quantum chemistry demonstrations in medicine, and it is worth naming plainly rather than treating as a defect. Nobody validates a new instrument on an unknown sample. The question a hospital pharmacology group or a pharmaceutical modeling team should ask is what the next experiment is: at what system size does the classical benchmark become unavailable, and does the quantum pipeline still agree with anything at that point. The release does not say.
What the announcement leaves out: no paper, no error bars, no runtime
The release names no preprint, no journal submission and no repository. There is no shot count, no reported statistical uncertainty on the 0.5 kcal/mol figure, no wall-clock or queue time for the Forte job, and no identification of which classical method produced the benchmark. For a result whose entire content is a numerical agreement, the absence of an uncertainty estimate is the most consequential gap, because 0.5 kcal/mol against a one kcal/mol threshold is a margin that a modest error bar would swallow.
Two other statements sit outside the demonstration and should be read as product description rather than result. The claim that Promethium runs demanding calculations up to 20 times faster than conventional CPU-based density functional theory platforms concerns GPU-accelerated classical chemistry and has nothing to do with the quantum machine. The suggestion that better binding energies at iron sites "could improve candidate ranking and help identify metabolic risks earlier" is a statement about what the technology might eventually support, and the conditional is the company's own.
How Quentir Reads It
The useful signal in this announcement is not the energy figure. It is that a quantum chemistry vendor reports a second run of the same workflow on another architecture, which moves the conversation away from any single machine and toward the software layer that decides what the machine is asked to do. Portability is what the company claims, and it is worth separating from comparability: the August disclosure carried no numerical result that today's 0.5 kcal/mol figure could be set against, so no one outside the two companies can yet compare the two runs. For a hospital or a pharmaceutical buyer, that matters more than qubit counts, because the active-space selection, the embedding scheme and the error mitigation are where the scientific risk lives, and those travel with the vendor rather than with the processor.
Two questions are worth putting to any hybrid quantum chemistry result of this kind, and they are this Monitor's own rather than a published checklist. First, who chose the active space, and is that choice reproducible by someone outside the vendor. Second, what is the uncertainty on the headline number, expressed the way an experimentalist would express it. On the evidence released on 1 September 2026, both are open.
The milestone to watch is a P450 result on a human enzyme with a therapeutic question attached, run at a system size where correlated classical chemistry becomes impractical, with an uncertainty estimate and a public method description. Until then, the sensible reading of this demonstration is the modest one: a workflow the company ran on IBM hardware in August has now been run on trapped ions in September, on a machine well characterized enough for that second run to mean something. That is a step on a ladder whose upper rungs are still empty, and steps on ladders are how instruments get built.
Sources
Primary source: QC Ware and IonQ, "QC Ware and IonQ Demonstrate High-Precision Hybrid Quantum Workflow for Drug Discovery," PR Newswire, 1 September 2026, carrying the quotations from Kin-Joe Sham, co-founder and chief operating officer of QC Ware, and Scott Millard, chief business officer of IonQ, and from which the Promethium and IonQ Forte pairing via Amazon Braket, the heme active site of cytochrome P450nor as the target, the 0.5 kcal/mol agreement with classical benchmarks at roughly four percent, the one kcal/mol chemical-accuracy threshold, the claim of more than double the accuracy of the standard classical mean-field method, the 115-atom model with over 1,000 molecular orbitals, the automatic reduction to a four-orbital active space mapped onto eight qubits, the single measurement basis, the classical computation of final interaction energies, the Amazon Web Services cloud compute credits and the statement that Promethium runs selected workloads up to 20 times faster than conventional CPU-based density functional theory platforms are all taken. The biology of the modeled enzyme comes from Hirofumi Shoun, Shinya Fushinobu, Li Jiang, Sang-Wan Kim and Takayoshi Wakagi, "Fungal denitrification and nitric oxide reductase cytochrome P450nor," Philosophical Transactions of the Royal Society B, volume 367, issue 1593, 2012, doi 10.1098/rstb.2011.0335, which identifies P450nor as a fungal CYP55 nitric oxide reductase, characterized in Fusarium oxysporum among other fungi, that reduces nitric oxide to nitrous oxide and receives electrons directly from reduced nicotinamide adenine dinucleotide, NADH. The company release names no organism and no accession for the modeled enzyme. The 36 qubits, all-to-all connectivity, 0.02 percent one-qubit gate error, 0.4 percent two-qubit gate error and 0.5 percent state preparation and measurement error attributed to IonQ Forte are taken from IonQ's own published system page. The identification of this demonstration as the IonQ follow-up to the IBM Heron run, and the four disclosures today's release adds to that earlier account, rest on QC Ware's own PR Newswire release of the Promethium and IBM Quantum demonstration and on this Monitor's reading of it published on 7 August 2026. The reading of the mean-field comparison as the weaker of two comparators at an open-shell iron center, the observation that a validation against an existing classical benchmark bounds what the result can establish, the note on the absent shot count, uncertainty estimate, runtime and benchmark-method identification, the warning that automated active-space selection hides its own errors, and the two diligence questions put to a hybrid quantum chemistry result are this Monitor's own and are attributed neither to the companies nor to the cited authors.