STRC Research

Phase 7I — APBS off-target rescore (CONHOH lead vs CONHOMe Tier-1)

4 min read

Phase 7I APBS off-target rescore — head-swap selectivity test

Two runs, one finding

Run A (Phase 8g_v2, pre-existing): v5.2__aq3__adamantyl__CONHOH__-Cl (mech-anchor reference) vs 4 cochlear off-targets + STRC E1659A K1141 reference. Result file: artifacts/phase8g_v2_apbs_offtarget.json (was unanalyzed before this session).

Run B (Phase 7I, this session): v5.2__aq3__adamantyl__CONHOMe__-Cl (Tier-1 #2, Boltz Δ +0.066) on the same panel. Script: scripts/phase7i_apbs_offtarget_conhome.py. Result file: apbs_offtarget_conhome.json.

Combined results

TargetCONHOH φ (kT/e)CONHOMe φ (kT/e)Δ (CONHOMe − CONHOH)Selectivity vs STRC
STRC E1659A K1141 (on-target)+2.51+2.48−0.03ref
KCNQ4 (7BYM)+0.91+1.12+0.21🟢 selective
TRPM4 (8RD9)+9.36+6.12−3.24🔴 anti-selective (worse)
TMEM16A (7ZK3, 6 metals)+4.90+4.27−0.63🔴 anti-selective
Cx50 (8QOJ via Cx36 paralog)−3.02−3.55−0.53🟢 strongly selective

φ = mean potential at lead’s hydroxamate / methyl-ester anion-O atom positions (the canonical “anion site” for the formal-anion Coulomb pull, identical method between runs).

Headline findings

  1. The CONHOH lead FAILS the +2 kT/e off-target selectivity gate as defined in the Phase 8g_v2 protocol. TRPM4 anion-O φ (+9.36) is +6.85 kT/e higher than STRC (+2.51). TMEM16A also exceeds STRC by +2.39 kT/e. This Phase 8g_v2 result was sitting unanalyzed in JSON since 2026-04-25. It is a paper-claim-load-bearing finding that has now been surfaced.

  2. CONHOMe head-swap REDUCES off-target electrostatic affinity without losing on-target. TRPM4 drops by 3.24 kT/e (the largest swing). STRC is essentially unchanged (−0.03 kT/e, within numerical noise). This means the head swap is a partial selectivity rescue:

    • On TRPM4: gap shrinks from +6.85 to +3.64 kT/e (still anti-selective but materially better).
    • On TMEM16A: gap shrinks from +2.39 to +1.79 kT/e (now under the +2 kT/e gate; passes strict-margin test on TMEM16A specifically).
    • On Cx50 / KCNQ4: gap unchanged or marginally improved.

  3. CONHOMe Tier-1 still FAILS the strict +2 kT/e all-targets gate because TRPM4 +6.12 kT/e exceeds STRC +2.48 by 3.64 kT/e. The head swap is mechanistically helpful but does not resolve the TRPM4 specifically.

  4. The phosphonate Tier-1 candidates are not yet APBS-tested because off-target Vina poses do not exist for them. Phosphonate carries −2 charge; the prior expectation is that it will worsen off-target electrostatic attraction (especially at metal-coordinating sites), but that needs measurement, not assumption. Queued as P0 light: aq3__1-indanyl__phosphonate__-CF3, aq3__4-F-biphenyl__phosphonate__-CN, plus the second CONHOMe variant aq3__adamantyl__CONHOMe__-CN.

What this changes about the lead picture

  • Phase 5k mech-4 holds — the +5.99 kT/e MUT-pocket attractor at K1141 NZ is real (decomposed-electrostatics, p=6.9e-12).
  • But the on-target attractor is not unique to STRC in the cochlear off-target panel: TRPM4 has a stronger anion-O attractor (+9.36) and TMEM16A has a comparable one (+4.90). The CONHOH-anion sees these too.
  • The on-target plus mut-prefer signal is what gives STRC its selectivity edge in principle, not the magnitude of the on-target attractor. Phase 7I shows mut-prefer lives on the head group: CONHOMe and phosphonate carry the Boltz-2 mut-prefer signal; CONHOH does not.
  • Lead picture for the paper:
    • Mech-anchor (§3): aq3__adamantyl__CONHOH__-Cl — load-bearing for Phase 5k formal-anion pocket attraction proof.
    • Lead candidate (§5): pivot to aq3__adamantyl__CONHOMe__-Cl — passes ADMET, salt-bridge K1141 NZ contact (3.89 Å), Boltz mut-prefer (+0.066), partial off-target rescue (−3.24 kT/e on TRPM4).
    • Selectivity caveat (§7): TRPM4 + TMEM16A residual electrostatic attraction is an outstanding question. Mitigation paths: (a) Boltz-2 + Vina dG analysis on the off-target binding modes — does the lead actually bind TRPM4 even with electrostatic pull? Or does the binding pocket lack the geometric fit? (b) Phosphonate Tier-1 might worsen this; (c) Imidazole / triazole bioisosteres (DR4 suggestions) that retain mut-prefer geometry without high formal-anion potential.

Note on the off-target ranking

For the CONHOMe lead, the φ ranking is: TRPM4 (+6.12) > TMEM16A (+4.27) > STRC (+2.48) > KCNQ4 (+1.12) > Cx50 (−3.55). KCNQ4 and Cx50 are cleanly selective. TRPM4 is the dominant anti-selectivity concern; TMEM16A is now within strict gate margin.

Connections

Graph View