August 2026 · Discovery · Verified architecture claim

KINASE SELECTIVITY 9×17

Selective versus promiscuous kinase inhibitors on IBM Heron r2 · ibm_kingston · 156 qubits. Pre-registered pair claim and architecture gap, scored PASS on raw QPC readout.

Plain English

What this case is: A kinase inhibitor is a drug that turns a protein switch (a kinase) off. The useful ones hit the disease target. The risky ones also hit other kinases, including cell-cycle and safety-related ones. Medicinal chemists constantly weigh: hit the intended target, spare the collateral kinases. This desk uses one complete public map (Davis 2011 Kd) of nine well-known inhibitors against seventeen kinases.
Why classical methods split: If you simply average how hard each drug hits all seventeen kinases, dasatinib looks best. It is potent and promiscuous. If you first require on-target activity and then spare AURKA, CDK2, and JAK2, imatinib looks best. It is more selective. Those two recipes disagree. That is the live chemistry question.
What we expected (locked before the run): On one coupled quantum job, with intended hits, collateral hits, and safety kinases sitting together: imatinib must outrank dasatinib, and the raw junction gap versus a no-cross control must be at least 0.05.
What QPC computed: One 156-qubit architecture job on ibm_kingston (Heron r2) · 4096 shots · raw counts primary · public job ID below.
Headline result: PASS: imatinib 0.610 > dasatinib 0.547. Raw junction gap 0.197 (need ≥ 0.05). The locked pair and the architecture gate both held.
What this delivers: A verified finding on a frozen public Kd table: when those logics co-reside, QPC sided with selective imatinib over promiscuous dasatinib. Architecture research on a public benchmark anyone can re-run.

Verdict

PASS. KIN-SEL-E1 and KIN-SEL-A1 verified

QPC ranked imatinib above dasatinib on coupled raw lead, against equal-weight mean potency on the same Davis panel. The raw junction gap was 0.197 versus the no-cross control.

PASS

Claim verdict

0.610

Imatinib lead

0.547

Dasatinib lead

0.197

Raw junction gap

Why this counts as a real result

Same question for one chemist: selective versus promiscuous on one complete Kd matrix.
Nine named clinical kinase inhibitors. Seventeen named kinases. One complete Davis Kd matrix.
Expectation written and locked before IBM submit.
Classical methods disagree on the pair (flat potency average vs sequential spare-safety).
Success = both gates verified on raw coupled counts: pair order and junction gap.

Method	Order on the pair
Equal-weight mean potency	dasatinib > imatinib (dasatinib ranks 1st overall)
Sequential: hit intended, spare AURKA/CDK2/JAK2	imatinib > dasatinib (imatinib ranks 1st overall)
QPC coupled raw (this job)	imatinib > dasatinib. PASS

Full QPC ranking (Davis 9×17 panel)

Rank	Inhibitor	Lead
1	sunitinib	0.654
2	gefitinib	0.648
3	lapatinib	0.647
4	imatinib	0.610
5	erlotinib	0.592
6	sorafenib	0.585
7	nilotinib	0.583
8	dasatinib	0.547
9	midostaurin	0.543

The locked claim is the pair only (imatinib vs dasatinib). The table is the full coupled readout on the frozen favorability panel. Dasatinib, the equal-weight potency winner, is near the bottom.

Job record

Field	Value
Backend	IBM Heron r2 · `ibm_kingston` · 156 qubits
Job ID	`da632keaa69c739lgpag`
Shots	4096
Geometry	9 inhibitors × 17 kinases + 3 junctions = 156 qubits · junction_ryy · no_cross
Primary metrics	KIN-SEL-E1 pair on coupled raw lead; KIN-SEL-A1 raw junction gap ≥ 0.05
Aer (same lock)	imatinib 0.635 > dasatinib 0.574; gap 0.242
Source	Davis et al. Nat Biotechnol 2011 Kd (one complete matrix)
Label	Architecture research on a frozen public Kd table. IBM Heron r2 · ibm_kingston.

KINASE SELECTIVITY 9×17

Verdict

Why this counts as a real result

Classical contrast (same Davis Kd)

Full QPC ranking (Davis 9×17 panel)

Job record