August 2026 · Real-world pharma task · Verified claim

PDT THERAPY F2

Photodynamic therapy agent selection on IBM Kingston — a pre-registered multi-logic expectation, scored PASS on raw QPC readout. Not “the job ran.” A real-world result.

Plain English

What this task is: Choose among five therapy-only PDT drugs under three goals at once: efficacy, toxicity, and selectivity. Detection/guidance agents were excluded.
What we expected (locked before the run): talaporfin (Laserphyrin) must outrank padeliporfin (Tookad) on QPC multi-logic priority — because mono-efficacy says so, while a simple equal-weight average says the opposite.
What QPC computed: One coupled quantum job on ibm_kingston · 18 qubits · 4096 shots · raw counts primary · public job ID below.
Headline result: PASS — talaporfin priority 0.546 > padeliporfin 0.493. The claim was verified.
What we claim — and do not: Claim: this pre-registered ranking expectation held on QPC hardware. Do not claim: a clinical prescription, VQE chemistry, or that every PDT claim will pass.

Verdict

PASS — expectation F2-E1 verified

QPC ranked talaporfin_sodium above padeliporfin on coupled raw multi-logic priority, against equal-weight classical on the same proxies.

PASS

Claim verdict

0.546

Talaporfin priority

0.493

Padeliporfin priority

+0.053

Margin

Why this counts as a real result

Expectation written and locked before IBM submit.
Classical methods disagree on the pair (mono-efficacy vs equal-weight).
Success = claim verified, not job completion (QPC already has 30+ hardware proofs).
Architecture metrics on raw counts; resilience was diagnostic only.

Classical contrast (same proxies)

Method	Order on the pair
Mono-efficacy	talaporfin > padeliporfin
Equal-weight average	padeliporfin > talaporfin
QPC coupled raw (this job)	talaporfin > padeliporfin — PASS

Full QPC ranking (therapy-only panel)

Rank	Agent	Priority
1	verteporfin (Visudyne)	0.609
2	talaporfin_sodium (Laserphyrin)	0.546
3	padeliporfin (Tookad)	0.493
4	temoporfin (Foscan)	0.435
5	porfimer_sodium (Photofrin)	0.414

Job record

Field	Value
Backend	`ibm_kingston` (Heron · open plan)
Job ID	`da0bkvv2sl0c73bmtf2g`
Shots	4096
Geometry	5 agents × 3 logics + 3 junctions = 18 qubits
Primary metric	F2-E1 claim verdict on coupled raw priority
Honesty	Architecture research on frozen public proxies — not a prescribing tool

An earlier full-panel claim (E1: keep hexaminolevulinate out of top-3) scored FAIL — a real negative finding. Setup F2 is the therapy-only redesign that delivered this verified PASS.