Skip to content

ABCG2 Q141K pharmacological-chaperone re-screen (comp-047)

Question

Can this static-receptor docking configuration rank an FDA-approved molecule for Q141K ABCG2 trafficking rescue after known ABCG2 relationships are excluded?

The run attempted 135 molecules; 134 produced complete docking-score rows, and cyclosporine A did not. It compared a modeled region around residue 141 with a Walker-A box, then applied a separate ABCG2 exclusion layer. The scores are computational triage signals, not binding, folding, trafficking, or urate-transport measurements.

Verdict: inconclusive — no defensible docking-backed ranking

After both exclusion checks, the executable result contains 0 yes rows and 1 uncertain row: vorinostat. Rosuvastatin was the other original uncertain docking-tier row, but it is excluded because the FDA CRESTOR label identifies it as a BCRP substrate and the UniProt/DrugBank relationship set also flags it.

Vorinostat's margin is small and its position is not robust enough to make it a docking-backed wet-lab priority. It has a separate reason to appear in the validation assay: Basseville et al. reported increased Q141K ABCG2 expression, surface trafficking, and substrate efflux after vorinostat treatment (In Vitro; PMID 22472121). That phenotypic precedent does not establish direct binding to the modeled site or validate the docking row.

The result invalidates the ranking configuration. It does not invalidate the Q141K rescue route or establish that no druggable rescue site exists.

Why the ranking is not decision-usable

The modeled system does not represent the rescue mechanism

Q141K was modeled as a static side-chain substitution in an AlphaFold apo monomer. A pharmacological chaperone acts on a folding-competent state or ensemble. This experiment does not model a folding intermediate, folding free-energy change, mutant-selective stabilization, the ATP-bound ABCG2 dimer, or intracellular exposure at the folding compartment.

The residue-141 box is a local structural region, not an experimentally validated pocket. The Walker-A box is neither the physiological composite ATP site nor the transmembrane substrate cavity. Their score difference therefore does not establish a selective fold-site interaction.

The base ordering is unstable

The recorded sensitivity run re-docked only the Q141K fold-site box. Its limited panel used x +2 Å, x -2 Å, y +2 Å, one +3 Å xyz diagonal, two box sizes, two alternate Vina seeds, and a neutral-ligand condition. It did not test y -2 Å, either z direction, the Walker-A box, or the complete executable margin rule. Across its non-base perturbations, 2–7 of the eight tracked candidate positions changed. That makes the base-run fold ordering descriptive for this setup, not a robust shortlist. It does not prove that a pocket is absent or establish robustness of the complete classification.

There is no ABCG2 chaperone positive control

The four CFTR correctors are cross-protein chaperone mechanism comparators. None reached the executable tier. Because they are not validated ABCG2 fold-site binders, this is a setup diagnostic—not a sensitivity estimate for ABCG2 chaperones.

The curated ABCG2 inhibitor/substrate controls remain excluded. That shows the declared exclusion layer functions for those controls; it does not validate the fold-site ranking.

Evidence-axis boundary

  • Axis 1: frozen Vina scores and the original transparent docking tier.
  • Axis 2a: bounded ChEMBL ABCG2 activity checks. No ChEMBL record is not evidence of no transporter relationship.
  • Axis 2b: UniProt-exposed DrugBank ABCG2 relationship flags, used as conservative exclusions. A flag is not relabeled as proof that every listed molecule is a substrate.
  • Independent substrate evidence: the FDA label establishes rosuvastatin's BCRP-substrate status.

Axis 2 can exclude a row. It cannot promote a survivor or establish pharmacological rescue.

Receptor-integrity result

The frozen WT and Q141K PDB/PDBQT files and grid boxes pass exact SHA-256, atom/residue-count, residue-141, mutation-scope, and geometry checks. The clean WT and Q141K structures differ only at residue 141.

The verifier records one declared preparation warning: Open Babel renamed terminal SER655 to UNK in both PDBQT files. This symmetric warning does not repair the missing score-to-receptor provenance of the historical docking run and does not justify retroactive re-docking for this inconclusive result.

Decision and next observation

  • Do not use the comp-032 class-prior list or COMP-047 base-score ordering to choose a compound.
  • Preserve the direct-chaperone route as an unvalidated hypothesis.
  • Treat vorinostat, romidepsin, panobinostat, valproate, and tubastatin according to their independent Basseville control roles—not according to COMP-047 rank.
  • Resolve the route in validation experiment §1.22: Q141K surface trafficking, ABCG2-attributed urate flux, direct inhibition, intracellular exposure, viability, and barrier integrity.

Another pass through the same static docking configuration is not the next experiment. A future folding-ensemble or ΔΔG model would be a new computational experiment with its own lifecycle.

Artifact

Experiment README · machine result · summary · control read-out · receptor verification