Pharmacochaperone Residence Time Criterion
Distilled from DR3 — CRO Wet-Lab Vendor Menu §I.4. The single most important mechanistic claim that H01 has not yet quantified.
The principle
For a pharmacochaperone — unlike a competitive inhibitor — equilibrium K_d is not the success metric. The compound must remain bound to the misfolded mutant long enough for the protein to traverse ER quality control (calnexin/calreticulin cycle, ERAD timers, COPII vesicle packaging). Rough order-of-magnitude: ER-to-Golgi residence on ~tens of minutes to hours.
Operative criterion: slow k_off — equivalently, long target residence time τ = 1/k_off — beats nanomolar K_d achieved via fast-on/fast-off binding. A 1 µM K_d compound with k_off = 10⁻⁴ s⁻¹ (τ ≈ 3 h) outperforms a 10 nM compound with k_off = 1 s⁻¹ (τ ≈ 1 s).
This is the same logic Copeland 2006 laid out for sustained pharmacology generally; for chaperone rescue it is harder-edged because the chaperone “job” has a temporal floor (ER transit time), not just a clinical PK ceiling.
Why H01 currently has a blind spot here
Every H01 affinity number to date is equilibrium-flavored:
- Vina ΔG → K_d via Δ G = RT ln K_d
- MM-PBSA / MM-GBSA in STRC h01 Phase 5 MD Ensemble Rescoring 2026-04-23 → ΔG_bind ensemble averages
- Phase 5k matched-ensemble APBS → ΔG_elec preference
- Phase 8h-lite Vina pose-stability across 20 snapshots → static spatial persistence
None of these touch k_off. The lead v5.2__aq3__adamantyl__CONHOH__-Cl could be a 100 nM thermodynamic binder with a 100 ms residence time — and would still fail mechanistically as a chaperone.
Computational paths to k_off
In rough order of cost vs. fidelity on local Mac silicon:
- τRAMD (random acceleration MD) — applies stochastic perturbation force, measures unbinding times across replicas, ranks compounds by relative residence time. ~10-20 replicas × 5-10 ns each per ligand. Not absolute k_off, but rank-correlates well per Kokh 2018. Fits Phase 5d/5e infrastructure.
- Metadynamics along an unbinding CV — well-tempered metadynamics with the ligand–pocket distance + pose orientation as collective variables. Returns absolute ΔG along the unbinding path, which combined with TST gives k_off. ~10× more compute than τRAMD; requires CV choice review.
- Markov state model from long unbiased MD — gold standard but needs hundreds of ns of unbiased trajectory. Off-budget unless cloud GPU.
Recommended next phase (Phase 5m candidate): τRAMD on the 14-compound v5.2 ADMET-clean shortlist + diflunisal positive control, on the K1141 fragment construct (per STRC K1141 Fragment Construct Strategy) once that AF3+MD ladder closes. Output = relative residence-time ranking. Reranks the v5.2 shortlist on the actual mechanistic axis, not on K_d alone.
Implication for the H01 paper claim
The current paper draft positions the lead on Vina ΔG + APBS + ADMET. Without a residence-time argument the mechanistic story is incomplete: a referee can ask “why won’t this binding equilibriate before ERAD fires?” and the answer right now is hand-waving toward Coulomb depth. τRAMD on the lead converts that into a quantitative response, even if absolute calibration is loose.
Connections
[part-of]h01 hub[refines]STRC Pharmacochaperone Virtual Screen E1659A[motivates-phase]STRC h01 Phase 5 MD Ensemble Rescoring 2026-04-23[depends-on]STRC K1141 Fragment Construct Strategy