Skip to content
Worked candidate · real output
From a whole-body mTORC1 run

This is what arrives.

One real molecule, unedited, exactly as the platform produced it — the structure, the two steps that build it, what it is likely to do in the body, and a plain account of what it is not.

GQ-259 · mTORC1 inhibitor✓ Pass
2D structure of GQ-259
COC1=C/C(=C\C(=O)C(=O)Nc2nc(O)c3cc(C)ccc3n2)OC(O)=C1
Identity
C₁₈H₁₅N₃O₆
Formula
369.3
MW
2.0
LogP
130.9 Ų
TPSA
3 / 8
HBD / HBA
4
Rot. bonds
2
Aromatic rings
OABSZHCXIYNYQS
InChIKey
Target · mTORC1Scope · Whole-bodyClass · SenomorphicRoute · 2-step, verified
Complete scoring
run of 2026-05-29 · scoring since revised
This molecule has been re-scored

Its original run published 1.00, normalised to its own batch. For two months we showed a ceiling of 0.54 or below instead, because nothing in the system could score one molecule on its own. That is now built, and the number is 0.35 — a measured value, every term of it reproducible. The bound we published was true, and it sat 55% above the answer.

Three figures moved and all three moved down. Pharmacophore fit was 0.98 on an axis we have since corrected; on the fixed axis it is 0.27. Selectivity was 6.7× and is 4.84×. Synthetic accessibility reads 0.79 rather than 0.74 because the score uses a blend of the SA heuristic and a retrosynthetic score, and only the first half was ever shown. Structure, route and ADMET are unaffected by any of it.

Two things we had not said. This molecule fires three structural alerts — it carries an alpha-keto-amide, which is reactive — and the 15% penalty for them is already inside the 0.35. And it is scored whole-body: asked for a brain-penetrant candidate it is rejected, because its predicted blood-brain-barrier permeability falls below the floor we require for a CNS target. This dossier stands as a worked example of the output and of the route. It is not a recommendation to make this compound.

Potency & drug-likeness
Overall0.35
re-scored, 0.337–0.345
Target activity0.23
low, expected
QED0.55
drug-like
Synthetic accessibility0.79
SA 0.74 + RAScore 0.84
Pharmacophore fit0.27
weak, corrected axis
Shape similarity0.53
moderate, atom-order band
Selectivity (SSR)4.84×
ratio, was 6.7×
Drug-likeness1.00
no Ro5 violations
Structural alerts3
alpha-keto-amide, x0.85 applied
Geroscience priors
Signature reversal (ECR)0.57
prior only
Hallmarks localized3 / 12
target-level prior
Drug classSenomorphic
Geroprotector likeness0.15
novel, unlike known
López-Otín 12-hallmark localization3 / 12 ≥ 0.5
GenomicTelomereEpigeneticProteostasisAutophagyNutrientMitochondriaSenescenceStem cellIntercellularInflammationDysbiosis
What it engages

A nutrient-sensing / autophagy senomorphic.

Deregulated nutrient sensing
0.95
Disabled autophagy
0.90
Loss of proteostasis
0.70
Cellular senescence
0.45
Altered intercellular comm.
0.45
Mitochondrial dysfunction
0.40
Stem-cell exhaustion
0.40
Phenotype · across 10 tissues

The aging-signature readout.

0.57
signature reversal
3 / 12
hallmarks localized ≥ 0.5

A mechanistic structural prior, disclosed, not a lifespan claim (see the methods note). It tells you where the candidate is predicted to move the aging signature.

Per-tissue aging-signature localization · relative
Thymus
100
derived-from-immune · no thymus data behind this clock
CNS / Brain
75
Hypothalamus
69
Vascular
62
Heart
60
Immune (PBMC)
57
Gut
57
Skeletal muscle
53
Liver
41
Adipose
22

Relative, within this candidate -- not a claim that any tissue gets younger. The thymus row is greyed because its clock is built from the immune clock by hand and contains no thymus measurement; it is shown rather than hidden so the ranking is not quietly reordered to flatter us, but it is not evidence.

Synthesis · verified by construction

A two-step route, because it was built from reactions, not guessed.

01 · Scaffold
Carboxylic-acid template
design-library core
02 · Step 1 · esterification
Pyranone oxo-ester (BB1)
USPTO acyl/ester template
03 · Step 2 · amide coupling
Amino-quinazoline (BB2)
HATU / DIPEA

Each atom is traced to its origin synthon, so the route is true attribution from the recipe, not post-hoc explanation. Building blocks here are drawn from the design library; for a live program we map candidates to catalogue building blocks carrying supplier part numbers before hand-off. A catalogue match is not a stock check — we hold no supplier account, so we cannot confirm anything is available to order.

ADMET · ML-predicted
hERG0.21
low cardiotox risk
Ames0.24
low mutagenicity
HIA0.93
high oral absorption
LogP (ML)1.17
well within range
Solubility−3.9
adequate (log mol/L)
Caco-2−5.16
borderline permeability
BBB0.06
peripheral, fine for a whole-body target
Verdict
6 of 7 clean
Caco-2 borderline, flagged, not hidden.
In the interest of honesty
  • Novel chemotype: similarity to known geroprotectors is low (≈0.15), a freedom-to-operate signal, not a red flag, but it also means there is no close precedent.
  • The phenotype and clock-reversal figures are mechanistic priors, disclosed, not validated against lifespan.
  • This molecule comes from an earlier run that predates our binding-plausibility and lifespan-prior scoring, so those are not reported here; a re-score would populate them.
  • GQ-259 is a computational hypothesis for medicinal chemistry, not an approval-ready compound. It requires synthesis and wet-lab validation.