Skip to content

comp-047 — Summary

Frozen docking run date: 2026-07-14

Method: static-receptor AutoDock Vina at a modeled residue-141 region and Walker-A comparison box, followed by ChEMBL inhibition and UniProt/DrugBank relationship exclusion.

Vina: seed 20260714, exhaustiveness 8, cpu 4; 135 attempted and 134 complete docking-score rows; incomplete: cyclosporine_a.

VERDICT: INCONCLUSIVE — NO DEFENSIBLE DOCKING-BACKED RANKING

Executable rule output after both Axis-2 checks: 0 yes, 1 uncertain. An executable row is not a wet-lab priority: the screen has no validated ABCG2 chaperone positive control, uses a static conformation, and produced unstable fold-site rankings under the recorded perturbations.

Executable marginal rows (not wet-lab priorities)

rank molecule class fold@Q141K Walker A margin Q141K−WT proxy ChEMBL activity DrugBank relationship tier
1 vorinostat HDAC inhibitor -6.19 -5.64 0.55 0.23 no bounded record no uncertain

Base-run fold-site scores (descriptive, not a robust ranking)

The table records the strongest scores in the original box. It is not a fallback shortlist: the sensitivity artifact shows material rank changes under box, seed, and protonation perturbations, and Axis 2 remains an exclusion layer rather than evidence of chaperone activity.

rank molecule class fold@Q141K Walker A margin excluded by Axis 2?
1 avacopan C5aR1 antagonist -7.82 -7.54 0.28 no
2 lumacaftor CFTR corrector -7.35 -8.77 -1.42 no
3 rosuvastatin Statin -6.80 -6.26 0.54 yes
4 ver155008 Hsp70 inhibitor (research) -6.46 -7.60 -1.14 no
5 spiperone Antipsychotic / receptor probe -6.45 -7.64 -1.19 no
6 mcc950 NLRP3 inhibitor (research) -6.38 -7.32 -0.94 no
7 novobiocin n/a -6.36 -6.61 -0.25 yes
8 elacridar P-gp/ABCG2 inhibitor -6.30 -8.23 -1.94 yes
9 glycerol_phenylbutyrate Ammonia scavenger / chaperone -6.20 -5.94 0.26 no
10 vorinostat HDAC inhibitor -6.19 -5.64 0.55 no
11 tariquidar P-gp/ABCG2 inhibitor -6.16 -7.75 -1.59 yes
12 sorafenib Multikinase TKI -6.11 -7.14 -1.03 yes
13 ezetimibe Cholesterol absorption inh -6.06 -6.23 -0.17 yes
14 lisinopril ACE inhibitor -6.06 -6.69 -0.63 no
15 gefitinib EGFR TKI / ABCG2 substrate -5.95 -7.18 -1.23 yes

Walker-A comparison diagnostic: scores span -8.77..-3.51 (median -6.09); modeled fold-site scores span -7.82..-1.85 (median -4.86). The substantial overlap makes the margin rule non-discriminating in this configuration; it does not establish a selective fold-site interaction.

Sensitivity diagnostic: the recorded panel re-docked only the Q141K fold-site box using x +2 Å, x -2 Å, y +2 Å, a +3 Å xyz diagonal, two box sizes, two alternate seeds, and a neutral-ligand condition. It did not test y -2 Å, either z direction, the Walker-A box, or the complete margin rule. Within that limited panel, 2–7 of the eight tracked candidate positions changed. The base-run fold ranking is therefore not treated as robust; this diagnostic does not establish robustness of the executable classification.

Interpretation boundary

  • Rosuvastatin is removed from the executable shortlist because it is independently identified as a BCRP substrate and is also present in the UniProt/DrugBank ABCG2 relationship set.
  • Vorinostat is the sole marginal executable row. Its direct Q141K rescue precedent is phenotypic and independent of this docking result; it does not validate the modeled pocket or make the docking row a wet-lab priority.
  • Failure to recover the CFTR comparators is a setup-specific diagnostic, not evidence that ABCG2 cannot be pharmacologically rescued.
  • The decisive next observation is the registered Q141K surface-trafficking + urate-flux + ABCG2-inhibition counterscreen in validation experiment §1.22, not another pass through the same docking configuration.

Load-bearing limitations

  • Q141K is a static side-chain substitution, not a folding-ΔΔG model.
  • A folding intermediate and mutant-selective stabilization are not modeled.
  • The receptor is an apo monomer; the Walker-A box is not the physiological composite ATP site or the transmembrane substrate cavity.
  • Vina scores and close margins are not binding-affinity measurements.
  • Exposure at the intracellular folding compartment is not modeled.

See controls.md, sensitivity.json, receptor_verification.json, and the README.