- When
- Monday 14 September 2026, 15:00–16:30 EDT
- Where
- Room 715A, Metro Toronto Convention Centre
- Format
- 90 minutes. Discussion, not talks.
- Moderator
- Alexander Chernov
- Slides
- doi:10.5281/zenodo.22841353
A quantum computation that is useful is not automatically a computation a regulated pipeline can use. Those are two separate thresholds, and there is not yet a mature evidence interface connecting them.
Most life-science quantum demonstrations are still checked against an established classical method. That is the right way to start. It is not evidence that anything survives contact with an industry that has to defend its results.
The tempting version of this argument — that regulated science needs determinism and quantum machines are stochastic — is wrong, and a room full of people who run validated pipelines would say so. Biopharma has handled biological variability, analytical error and calibration uncertainty for decades. What it demands is not identical numbers. It is controlled uncertainty and attributable evidence.
So the real question is harder: what evidence makes a stochastic, hardware-dependent computation reproducible enough, attributable enough and controlled enough for its intended use?
Three questions, in this order
- What evidence must accompany a quantum result before a regulated workflow can even evaluate it?
- Which biological use cases are credible under each regime — investigable on NISQ now, dependent on fault tolerance, or lacking a defensible quantum rationale at all?
- Which governance requirements are regime-independent — true whether the device is noisy or fault-tolerant?
Everything else on the agenda is in service of these three.
Four seed positions, deliberately incompatible
They exist so the room has something to disagree with immediately. You may adopt one, amend one, or reject all four.
The bottleneck is not qubits, it is evidence.
Grant a useful quantum machine tomorrow: regulated biology still could not consume it. The gap is a governance and instrumentation gap — platforms expose pieces of the required evidence, but the complete, durable evidence interface a governed workflow needs is still immature, and the conventions for it are being set now by whoever ships first.
The session succeeds if the room dismantles it.
This session is fifteen years premature.
There is no biological problem where quantum beats classical this decade. Worrying about the regulatory provenance of current demonstrations is like drafting flight-safety rules for a paper airplane. Build the fault-tolerant machine first.
Only quantum chemistry is real; the rest should stop.
The one defensible near-term target is molecular simulation, where the problem is natively quantum. Genomics-as-QUBO, quantum classifiers over biological feature vectors and bioprocess optimization take problems classical computing already handles well, encode them awkwardly, and run them worse.
The interface is the whole problem, not the advantage.
Quantum subroutines will sit inside classical loops. The engineering question is the boundary: what crosses it, what guarantees hold across it, and what happens when the quantum side returns a distribution to a caller that expected a number.
Four artifacts, offered to be attacked
All modest by construction. Their value is not that they are impressive — it is that they are concrete enough to argue with.
-
A gate that refuses quantum results it cannot certify
Runs entirely in a browser tab, no backend and no hardware, so you can recompute its decisions yourself. It scores itself against the session’s own readiness checklist and fails four of eight rows on purpose. qubit.agenticdatasets.org
-
A published quantum-genomics method, reproduced
QuASeR’s QAOA de novo assembly on PennyLane. Nine qubits, a five-base contig, verified against the brute-force optimum — which is exactly why it proves nothing about advantage.
-
A trained quantum classifier, downstream
It does not beat a linear classifier and no claim is made that it does. It is included because it is governed on the same terms as the stage above it.
-
A regulated decision, from the other side of the wall
A mechanistic fed-batch CHO bioreactor twin under a contract of in-process limits tied to a Critical Quality Attribute, emitting a decision artifact. No quantum step in it at all. It is the yardstick.
Who is moderating
Alexander Chernov
Associate Principal Data Engineer, Data Science and Modeling — BioPharmaceuticals R&D, AstraZeneca, Greater Toronto Area
He works on the control planes behind regulated biopharmaceutical science: the systems deciding what may be executed, and what a computed number must prove before anyone may act on it. He wants to know whether the quantum community grasps that this is the bar.
He has reproduced a published quantum-genomics method — QuASeR’s QAOA-based de novo assembly — on the PennyLane stack, trained a variational quantum classifier downstream of it, and placed both behind a descriptor-and-gateway control plane with resource budgets, admission control and an event record. He separately built a mechanistic digital twin of a fed-batch CHO bioreactor under a contract of in-process limits tied to a Critical Quality Attribute. He is an associate editor for the IEEE Toronto Section newsletter, and in 2026 has had work accepted at IEEE CCECE, EMBC, CBS and BigDataService, spanning agentic data services, dataset descriptors, and policy-aware supervisory control for biomedical systems.
He comes as a systems engineer working inside biology, trained as a physicist, with an MSc in theoretical physics specializing in photoionization and confined atoms. He had not pursued quantum computing academically until recently. For a moderator on this question he regards that as an asset: he has no quantum result to defend, and every incentive to let the room reach the honest conclusion, including the unwelcome one.
Slides
The session was held on Monday 14 September 2026. The slides are published as a citable record under CC BY 4.0, except the portrait photograph.
Download the slides (PDF, 50 pages) Zenodo record
This edition carries the framing, the four seed positions, and five segments of supporting material — the certificate behind the live artifact, the QuASeR reproduction, the trained classifier, the antibody process and the CHO bioreactor model, and the strongest case against the session’s own premise — with every source. Speaker notes and the slides used to collect the room’s input are omitted.
Cite as: A. Chernov, “Is Quantum Ready for Biology? What It Would Take to Put a Quantum Step Inside a Regulated Biopharma Pipeline,” BOF-2099, IEEE Quantum Week 2026, Toronto, 14 Sep. 2026, doi:10.5281/zenodo.22841353.
Pre-read
The 15-page pre-read circulated before the session remains available, also under CC BY 4.0 except the portrait photograph. It is the shorter way in: the framing only, without the supporting segments.
Download the pre-read (PDF, 15 pages)
It carries the framing in full: where a regulated pipeline begins, what a computed result must carry before it may influence a regulated decision, what a quantum device actually returns, the three questions in order, the four artifacts, the four seed positions, and the sources behind every definition.