# SYNTHESIS REPORT — STAGE 2: GBA/Gαi3 interface + phospho-site / extreme C-terminus

Follow-on to Stage 1 (which found, on the disordered Girdin-CT ensemble, that no site reproduces the experimental Kd-vs-4-R-structure SAR above the docking-noise floor; the authors' homology "SH2" pocket passes the coarse SAR only via a non-pocket-specific ligand-size effect on a refuted fold; best-occupied blind site B01 overlaps the GBA/Gαi3 interface). Stage 2 tested the two deferred surfaces with the identical, twice-adversarially-reviewed apparatus (validated ligand series, canonical within-complex MDE estimator, primary-source Kd oracle, Signal A/B, 720-permutation multiple-comparison guard, honest-negative-is-a-pass, fine-ordering-never-claimed).

## Bottom line

Across **four surface classes now** — the disordered IDR ensemble and the authors' homology SH2 pocket (Stage 1), the **rigid experimental Gαi3 groove** and the **phospho-site / extreme C-terminus** (Stage 2) — **no surface reproduces the experimental SAR above its noise floor.** The Stage-1 NEGATIVE is therefore **not** merely an IDR-conformational-noise artifact: the ~0.6 kcal/mol Kd-vs-4-R dependence does not structure-based-dock-localize even on a rigid, real, experimentally-determined cavity. Two mechanistically real, positive findings do emerge, neither of which is an R-group SAR: (1) the complexes **occupy the Gαi3 GBA-binding groove strongly** (competitive block — consistent with 1a disrupting the Girdin–Gαi3 pulldown); (2) **anionic phospho patches electrostatically attract the +1 cationic complex** (a docking-score-blind effect), but as a general cation–anion attraction that does not reorder the series.

## Item 1 — Gαi3 Switch-II/α3 groove (rigid, real experimental cavity; 6MHF chain A)

**Verdict: rigorous NEGATIVE (PASS) — no robust pocket-specific SAR on the rigid cavity.** Docking the six complexes (1a-1e + R=H) into the GBA-binding groove of Gαi3 (competitive-block hypothesis):

- **Signal A (R essential) FAILS**: R=H (Irppy-bipy) is *not* the least-tight — it docks rank 4/6 (1e, 1b dock weaker).
- **Signal B is AT CHANCE**: its 720-permutation null pass-rate p=0.47 (≈coin-flip); the +0.082 kcal/mol gap only "clears" because the rigid single-conformer cavity's MDE collapsed to a **0.022 kcal/mol seed-reproducibility floor** (≈41× tighter than the Stage-1 IDR band of 0.7–1.1 kcal/mol). This collapsed floor is *not* the ~1 kcal/mol physical resolution of the scoring function, so the permutation null — not the sub-0.1 kcal/mol gap — is the operative guard. Both-signals null p=0.083.
- **Cross-surface caveat (mandatory):** the Kd *series* is complex↔Girdin-CT, not complex↔Gαi3. Cross-surface Spearman ρ=+0.37 (p=0.47, not significant); docking rank scrambled vs Kd. The only same-surface Gαi3 anchor (1a→Gαi3 Kd 15 µM) is a single point, not a series.
- **Positive mechanistic finding:** the complexes bind the Gαi3 groove **strongly and specifically** — groove best scores −9.1 to −10.6 kcal/mol vs a decoy/off-groove box −7.4 to −8.3 and the Girdin-GBA-helix control −6.0 to −7.1. So an Ir complex can occupy the site the GBA helix uses to engage Gαi3 → **competitive block**, the mechanism behind 1a's disruption of the Girdin–Gαi3 pulldown.

**Three readings of the negative are kept open:** (a) the SAR is below structure-based-docking resolution generally; (b) it was IDR-noise-specific in Stage 1 (a POSITIVE here would have favored this — it did not occur); (c) cross-surface mismatch (Girdin-CT SAR does not map onto Gαi3).

## Item 4 — phospho-site (~1705–1755) + extreme C-terminus (1815–1871)

### (4a) Blind SAR docking (apo) — rigorous NEGATIVE
Phospho region: 6 candidate sites screened, **0 clear both robust signals** (expected false positives = 0.000, 720-permutation). Best-occupied focus site B06; best R-bearing pose 1e −12.68 kcal/mol, 4-R in measured contact (2.83 Å) with an acidic residue (Glu1813) — geometrically admissible but not SAR-reproducing. The extreme C-terminus 1815–1871 (unsampled in Stage 1) was newly built (AF-full-length slice + MD-collapse to tractable compact variants) and docked; same NEGATIVE pattern. Same IDR behavior as Stage 1.

### (4b) Charge-sensitive phospho-electrostatic test (GFN2-xTB, geometry-matched)
Because smina/Vina scoring is **charge-independent**, the apo and phospho receptors dock identically → the phospho effect is invisible to the docking score and was probed only here, by geometry-matched apo-vs-phospho xTB ΔΔE_int (same docked pose; phosphate dianion −2 grafted on the identical geometry, 0.0 Å shared-atom deviation).

**Verdict: a real, docking-score-blind, directional electrostatic effect — but NOT an SAR.**
- Adding the phosphate makes the +1 complex's interaction energy markedly more favorable in **6/6 complexes at every mark** — mean ΔΔE ≈ **−61 kcal/mol** (pS1717), **−79** (pS1820), **−79** (pS1837) — i.e. the anionic phospho patch attracts the cation.
- **Magnitude is an unscreened gas-phase upper bound** (aqueous screening reduces it ~1–2 orders of magnitude; a truncated-pocket ALPB(water) number for a buried dianion is unreliable and is **not** claimed — see `solvent_sensitivity.json`). The *finding is the direction + its docking-blindness*, not a magnitude.
- **Phosphorylation does not change the rank-order** — a movement invisible to the charge-independent docking score, and *not* an R-group SAR (it is a generic cation–anion attraction, ~1/r by pose geometry). UniProt-backed marks pS1717/pS1820/pS1837; pY1744 (paper-only) modeled as a labeled secondary variant with the same qualitative result.

## Integrated overall conclusion (Stage 1 + Stage 2)

The experimental dependence of the binding constant on complex structure (1a 1.3 < 1d 1.5 < 1e 1.6 < 1c 2.6 < 1b 3.6 ≪ R=H >20 µM) is a **real, ~0.6 kcal/mol (substituted-series) / ≥1.6 kcal/mol (R=H) effect that is NOT localized in, or reproduced by, the geometry of any single binding site** at the resolution of rigid outer-sphere docking + GFN2-xTB rescoring — tested now across a disordered IDR ensemble, the authors' homology SH2 pocket, a rigid experimental Gαi3 cavity, and phospho/C-terminal regions. The authors' single-pose "explanation" is an over-interpretation: their pocket reproduces only the coarse size trend, non-specifically, on a fold six methods refute.

**What IS supported (mechanism, not SAR):** 1a binds the Girdin C-terminus at micromolar affinity (experimental, real) via distributed outer-sphere contacts; it can competitively occupy the Gαi3 GBA-binding groove (explaining the Girdin–Gαi3 pulldown disruption); and cationic complexes are electrostatically drawn to anionic phospho patches. The R-group dependence of Kd is best understood as a **small, distributed physicochemical effect of the 4-R substituent (size / lipophilicity / basicity / electrostatics) below the resolution of site-based structural scoring** — which is exactly why a correct, honest methodology reports it as unreproducible rather than fabricating a pose that "explains" it.

**The more-correct methodology delivered:** disorder-aware ensemble → per-conformer pocket catalogue → blind global docking of the full series → SAR rank-order validation against the fixed experimental Kd yardstick with a claim-granularity noise floor, pre-registered robust signals, and a multiple-comparison guard → physics rescoring → rigid-real-cavity control (Gαi3) → charge-sensitive rescoring for electrostatic (phospho) effects → orthogonal-data consistency. A rigorous negative is a valid, publishable result. This is the falsifiable, reproducible alternative to the authors' single unvalidated docking pose.

## Резюме (RU)

**Итог Stage 2 (и всей работы):** зависимость константы связывания от строения комплекса (SAR) **не воспроизводится структурным скорингом ни на одной из четырёх протестированных поверхностей** — разупорядоченный ансамбль и «SH2»-карман авторов (Stage 1), жёсткая реальная полость Gαi3 и фосфо/крайний C-конец (Stage 2). То есть отрицательный результат Stage 1 — **не** артефакт шума разупорядоченной мишени: SAR (~0.6 ккал/моль) не кодируется геометрией сайта на разрешении докинга. Модель авторов «объясняет» только грубый тренд по размеру лиганда, неспецифично, на опровергнутом фолде. **Что подтверждено механистически (но это не SAR):** комплексы сильно садятся в канавку Gαi3 (конкурентный блок → объясняет разрушение пул-дауна Girdin–Gαi3); анионные фосфо-участки электростатически притягивают катионный комплекс (эффект, невидимый для charge-independent докинга; направление устойчиво 6/6, но величина — незаэкранированная верхняя оценка, не заявляется; ранг не меняется → не SAR). Правильная методология (ансамбль → слепой докинг → SAR-валидация против фиксированного Kd + anti-gaming → physics-рескоринг → жёсткая-полость-контроль → charge-sensitive рескоринг) — честная, фальсифицируемая альтернатива одной невалидированной позе авторов.

## Caveats (honest)
- Cross-surface: the Gαi3-groove result is a different binding partner than the Kd oracle; framed as such throughout.
- The phospho electrostatic magnitude is an unscreened gas-phase upper bound — direction only is claimed, not magnitude; no SAR.
- The ~0.6 kcal/mol fine ordering is never claimed reproduced.
- The extreme-C-terminus compact variants are a docking-tractability sampling device, not a claim of a stable fold (the region is low-pLDDT disorder).
- All binding modes are computational hypotheses; no experimental Ir-complex + Girdin/Gαi3 co-structure exists.

**Evidence:** `stage2/gai3/` (GAI3_RESULTS.md, gai3_results.json), `stage2/phos/` + `stage2/cterm/` (BLIND_RESULTS.md, site/, rescore/PHOS_RESCORE.md + CTERM_RESCORE.md, solvent_sensitivity.json). Method: `stage2/report/METHODOLOGY_STAGE2.md`. Stage-1 report: `stage1/report/REPORT.md`. Companion (validated): Gαi3-DOCK + PHOS-DOCK sealed dual-lane; GAI3-REC/RESCORE + PHOS-REC sealed; the phospho-xTB (VAL-PHOS-003) compute completed on disk and is reported here (its Zenith worker finalization was blocked by an org-budget limit; the analysis + numbers are the worker's own committed outputs).