Maturity
3.5 / 10
Early. A VQE workflow exists and runs, but on a toy fragment with no classical comparator at the same problem size. Score is capped while one core advantage claim is needs-baseline.
Home / Sample briefs / Pharma
Pharma R&D — sample brief
Anatomy of a pharma brief.
An Illustrative end-to-end Veriqa evidence package on a synthetic molecular-design program. Same conservative discipline a real engagement produces.
00 · Brief metadata
Project Halberd-7 — strategic brief.
A synthetic molecular-design program submitted to Veriqa's Molecular Design Agent for evidence grading. The metadata below identifies the brief, the methodology version that produced it, and the scope of the engagement.
01 · Verdict
Require further diligence.
The verdict is decision-support, not a directive. It states what the evidence currently supports and what would change the grade.
Verdict
Require further diligence
QN-PH-0003
Quantum-chemistry binding-affinity claim is materially novel but unbenchmarked against the standard classical FEP+ comparator at this system size; needs wet-lab Kd confirmation and a head-to-head against MD / DFT before any downstream synthesis investment.
The team has submitted a credible VQE workflow on a small active-site fragment and reports a 0.7 kcal/mol affinity uplift on a synthetic congeneric series. Three of the five claim-discipline checks fail at this draft: the classical FEP+ comparator is named but not run at the stated problem size, no measured Kd is on file, and the quantum resource estimate assumes logical-qubit fidelity that current physical-qubit devices cannot deliver. The brief therefore caps maturity and blocks a Proceed verdict in code. A scoped 90-day plan (see §08) would resolve the open items.
02 · Decision frame
The five questions a pharma brief must answer.
Before grading the evidence, Veriqa forces a structured decision frame for the vertical. Every row of the evidence table is traceable back to one of these.
- Does the quantum approach beat MD / DFT / FEP+ on this case at this problem size? A claim of advantage requires a named classical comparator at the same heavy-atom count, basis set, and force-field treatment as the quantum run.
- At what cost — gate-time and qubit-count? A logical-qubit estimate at fault-tolerant fidelity is not interchangeable with a physical-qubit estimate on near-term hardware. Both numbers must be stated.
- At what TRL? Workflow demonstrated on a toy active-site fragment is not the same as a workflow deployable on a full protein–ligand complex. The maturity score follows the more conservative reading.
- Is the wet-lab confirmation pathway funded? An in-silico Kd is a hypothesis. A measured Kd from a defined assay is a result. The brief names the assay, the cohort, and the budget — or it does not.
- Is the IP defensible — composition vs. method? A patent on the molecule (composition of matter) carries different prosecution and infringement risk than a patent on the computational method. The brief separates them and names the filing strategy.
03 · Score gauges
Three scores, each with a reason.
Each score is a transparent, rule-based heuristic computed from features of the graded evidence — not a statistically validated model and not a sentiment score. Illustrative values for this synthetic brief.
Urgency
4.5 / 10
Medium. No regulatory deadline forcing a 90-day call, but the synthesis ramp decision is pending and a competing classical FEP+ run is already in flight inside the same group.
Hype / overclaim risk
6.8 / 10
Elevated. The press release oversold the result — "quantum advantage in drug discovery" appears in copy, but the supporting data is one VQE run on a fragment with no classical baseline and no wet-lab Kd.
04 · Classical-baseline comparison
VQE vs. the established classical comparators.
For an in-silico binding-affinity claim, the production-grade classical comparator is FEP+ (Free-Energy Perturbation, as implemented in Schrödinger's software). Density Functional Theory (DFT) and Molecular Dynamics (MD) cover adjacent system sizes. A VQE result that has not been measured against these is not an advantage claim — it is a method demonstration.
| Computational claim | Quantum method | Classical baseline | Problem size (heavy atoms) | Status |
|---|---|---|---|---|
| Binding-affinity uplift on congeneric series | VQE, hardware-efficient ansatz, simulator | FEP+ on the same series | ~38 (ligand) + active-site fragment | Needs-baseline — FEP+ run named in the plan but not executed at this problem size |
| Active-site electronic structure refinement | VQE on a 12-orbital active space | DFT (B3LYP / cc-pVTZ) | ~24 (fragment) | Run · DFT energy reported · VQE within 1.2 kcal/mol on simulator |
| Conformational sampling of ligand pose | (none claimed) | Classical MD, 200 ns trajectory | ~38 (ligand) + solvated complex | Run · classical only · used as input to the VQE step |
| Free-energy decomposition across the series | VQE-derived correction term | FEP / FEP+ free-energy perturbation | ~38 per ligand | Unverified — correction term applied but not validated against FEP at series scale |
| Resource estimate to reach production fidelity | VQE on fault-tolerant logical qubits | (not applicable — hardware claim) | Same active space | Needs-baseline — logical-qubit estimate cited; no physical-qubit estimate at current device fidelity |
"Schrödinger" appears here as the name of the software package whose FEP+ implementation is the production-grade classical comparator in the literature. It is not a competitor reference. Numbers and rows are Illustrative for the synthetic Project Halberd-7.
05 · Evidence table
Sub-claims, sources, confidence.
Every line carries a source ID and a confidence level. The verdict is a function of what is in this table — not an opinion layered on top.
| Sub-claim | Source | Confidence | Notes |
|---|---|---|---|
| VQE on the 12-orbital active space converges within 1.2 kcal/mol of DFT (B3LYP / cc-pVTZ). | SRC-01 — pre-print, arXiv:2602.08841 (synthetic) | Medium | Simulator only, no hardware run. Random seed not reported. |
| VQE step delivers a 0.7 kcal/mol binding-affinity uplift over the in-house docking baseline on the congeneric series. | SRC-02 — internal program memo (Halberd-7, undated) | Low | No FEP+ comparator at the same problem size. Flagged needs-baseline; caps maturity; blocks Proceed. |
| FEP+ is the production-grade classical comparator for relative binding affinity on congeneric series. | SRC-03 — Nature Reviews Drug Discovery, computational chemistry review (2024) | High | Establishes the comparator the team must run. |
| Public CASF-2016 benchmark scoring-function results bracket the expected error range for the target class. | SRC-04 — CASF-2016 public benchmark data | High | Used to sanity-check the claimed uplift magnitude against community-known scoring error. |
| Reference ligand in the congeneric series has a published Kd with a known assay protocol. | SRC-05 — ChEMBL row CHEMBL-XX-SYNTH (synthetic ID) | High | Anchors the wet-lab plan; uplift claims must be reproducible against this reference. |
| Cited resource estimate (≈140 logical qubits, ≈1.2×10⁹ T-gates) is consistent with published VQE active-space scaling. | SRC-06 — peer-reviewed VQE resource-estimate paper (2025) | Medium | Logical-qubit number. No physical-qubit translation at current device fidelity. |
| Press release headline ("quantum advantage in drug discovery") goes beyond the supporting data. | SRC-07 — program press release (2026-05-21) | Low | Drives the hype-risk score; recommended to retract or qualify before any external diligence cycle. |
| Composition-of-matter filing on the lead candidate is in preparation; method filing separate. | SRC-08 — internal IP counsel memo (2026-05-30) | Medium | Separation of composition vs. method is correct; claim language should track the conservative evidence log, not the press release. |
06 · Missing evidence
Exactly what is absent — and why it caps the verdict.
A conservative brief names the gaps. Each item below is what would have to be added, run, or measured to change the grade.
- No wet-lab Kd measurement yet. The in-silico uplift is a hypothesis until a defined binding assay produces a measured affinity on the lead candidate and on the reference ligand from SRC-05.
- No head-to-head FEP+ comparator on the same congeneric series. Until FEP+ is run at the stated problem size, the advantage claim cannot be recorded and the maturity score remains capped.
- No quantum resource estimate at production fidelity. The logical-qubit number does not translate directly to current physical-qubit devices. A physical-qubit estimate at named device fidelity must accompany the logical-qubit number.
- No IND-enabling tox plan. The synthesis ramp decision is downstream of, not parallel to, a credible early tox profile and a scoped IND-enabling plan. Neither is on file.
- No reproducibility note. Random seed, simulator versus hardware, ansatz parameters, and active-space selection rules are not documented at a level a second team could re-run.
- No third-party validation cohort. The single result is from one group on one instance. A second group reproducing the workflow on a related target would materially raise the maturity score.
07 · Validation plan
A 90-day path to a synthesis go / no-go.
A phased plan that ends in an explicit verdict revision on the synthesis ramp decision. Each step is scoped, owned, and gated; downstream steps do not start before the upstream gate is met.
-
Days 0–15 · Run the FEP+ baseline at problem size
Execute FEP+ on the same congeneric series at the same heavy-atom count and force-field treatment as the VQE run. Lock the comparator at the start, not after the result lands. Output: FEP+ relative affinities for the full series, comparable to the VQE numbers.
-
Days 10–30 · Reproducibility & resource estimate
Publish the VQE run with random seed, simulator/hardware split, ansatz parameters, and active-space selection rules. Add a physical-qubit resource estimate at named device fidelity alongside the existing logical-qubit estimate.
-
Days 20–60 · Wet-lab Kd on the reference ligand and lead candidate
Run the assay protocol from SRC-05 against the reference ligand (to anchor the assay) and the lead candidate (to test the predicted uplift). Output: measured Kd with confidence intervals and assay metadata.
-
Days 45–75 · Tighten the claim, retract or qualify the press release
Rewrite external copy to track the conservative evidence log. Coordinate with IP counsel so composition-of-matter and method filings carry only language the table supports.
-
Days 75–90 · Verdict revision and synthesis go / no-go
Re-run the brief with the new evidence. Maturity uncaps if and only if the FEP+ comparator clears, the wet-lab Kd tracks the predicted uplift within stated error, and reproducibility is documented. Verdict revision is the gate on the synthesis ramp decision.
08 · Reviewer status
Draft until the reviewer gate clears.
Reviewer status — draft. This brief stays in draft until an internal reviewer approves it; the software enforces the gate before delivery. Independent expert review by a computational chemist is recommended before any external diligence cycle or IP filing — we do not claim such independent experts currently staff sign-off.
Decision-support only. Veriqa's grading and verdicts assist your team's judgement; they are not a substitute for independent scientific, regulatory, legal, or investment review, and require your own verification. This brief is Illustrative for a synthetic fictional program and is not medical advice.
Path A engagement. Veriqa grades the evidence behind a design or claim. It does not design or run quantum circuits, integrate quantum hardware, or itself claim quantum advantage on your behalf.
Bring us your program.
A real Veriqa pharma brief on your design dossier, claim package, or vendor pitch — same conservative discipline, with your sources and your decision in scope.