# SYNTHESIS REPORT — How does Ir(III) carbene complex 1a bind the Girdin C‑terminus, and can the SAR be explained? (VAL‑REPORT‑001)

**Target.** Human Girdin / CCDC88A, UniProt **Q3V6T2** (1871 aa); experimental binding construct **His‑Girdin‑CT 1660–1870** (MST).
**Ligand.** Substitutionally‑inert octahedral iridium(III) carbene complex **1a** = [Ir(ppy)₂(4‑pyrrolidinyl‑iso‑BIPY‑NHC)]⁺, and its 4‑R congeners 1b–1e plus the R=H control.
**Source paper being tested.** Ruan M‑L *et al.* *Iridium(III) carbene complexes as potent girdin inhibitors against metastatic cancers.* PNAS 2024;**121**(25):e2316615121. doi:10.1073/pnas.2316615121 (PMC11194514). Full text `provenance/main_plain.txt`; SI `provenance/si/sapp.txt`.
**Companion document.** Method specification: [`stage1/report/METHODOLOGY.md`](METHODOLOGY.md) (VAL‑METH‑001).

---

## Bottom line (one paragraph)

Using a corrected, disorder‑aware, blindly‑searched, SAR‑validated protocol (Methodology §2), **no binding site on the Girdin C‑terminus reproduces the experimental structure–affinity relationship (SAR) above the method's own noise floor.** The blind whole‑surface search returns a rigorous **NEGATIVE** (4 candidate sites, **0 clear both robust SAR signals**, expected false positives ≈ 0), physics (GFN2‑xTB) rescoring of the best‑occupied site **does not confirm** it, and the only place the coarse SAR *does* reproduce — the authors' reproduced homology "SH2" pocket — reproduces it for a **non‑pocket‑specific shape/size reason** in a fold that six independent methods have refuted. The honest deliverable is therefore a **correct, reproducible, falsifiable methodology** plus an **honest hypothesis space**, not a proven binding site. The single best *structural* hypothesis carried forward (low confidence, explicitly not a validated lead) is a shallow, transient interaction near the **GBA/Gαi3 helix (canonical ≈1692–1744, site B01)** — consistent with the one experimentally‑validated function of this region (1a disrupts the Girdin–Gαi3 interaction), but *not* SAR‑reproducing. The fine intra‑amine ordering is **never claimed**.

---

## 0. Conventions (numbering + the read‑only oracle)

- **Numbering: canonical UniProt Q3V6T2. The paper numbers residues as canonical − 1** (ground truth `altrecept/numbering_manifest.json`, `evidence/verify_numbering.py`). Every residue below is canonical; the paper's "SH2 domain 1702–1835" is canonical **1703–1836**; the pEGFR mutagenesis set (paper K1722E/R1745/K1749E/Q1777E) is canonical **K1723 / R1746 / K1750 / Q1778**; the paper's conserved motifs G¹⁷⁴⁰XFXXR¹⁷⁴⁵ / pS¹⁷¹⁶XXF¹⁷¹⁹ are canonical **G1741…R1746 / pS1717…F1720**.
- **The experimental SAR is the fixed, read‑only yardstick, re‑extracted verbatim from the primary source** (`provenance/main_plain.txt` Fig. 7E/8E/7G + SI Fig. S20; cached with quotes in `stage1/dock_ctrl/kd_primary_source.json`, grep‑verified present in the article text):

  | complex | 4‑R group | Kd (µM) | primary figure |
  |---|---|---:|---|
  | 1a | pyrrolidinyl | **1.3** | Fig. 7E + SI S20 |
  | 1d | piperidinyl | **1.5** | Fig. 8E + SI S20 |
  | 1e | dimethylamino (NMe₂) | **1.6** | Fig. 8E + SI S20 |
  | 1c | morpholinyl | **2.6** | Fig. 8E + SI S20 |
  | 1b | tert‑butyl | **3.6** | Fig. 8E + SI S20 |
  | Irppy‑bipy (R=H) | none (plain 2,2′‑bipy) | **> 20** | Fig. 7G + SI S20 |

  Experimental rank (tightest → weakest): **1a < 1d < 1e < 1c < 1b ≪ R=H.**

- **Two robust signals** a correct model must rationalize, and **one non‑signal it must never claim:**
  - **Signal A — the 4‑R substituent is essential.** R=H is ≥ ~1.6 kcal/mol weaker than 1a (RT·ln(20/1.3) ≈ 1.62 kcal/mol at 298 K).
  - **Signal B — coarse chemotype:** small basic amine ring (1a/1d/1e) > bulky/ether/tert‑butyl (1c/1b).
  - **NOT a signal — the fine intra‑amine ordering** 1a<1d<1e<1c<1b spans only **~0.6 kcal/mol**, at/below docking‑score noise → **not reproducible, never claimed.**

---

## 1. The mission question

Ruan *et al.* place 1a in a single rigid homology‑modeled "SH2‑like" pocket and **never test that pose against the measured Kd series** — the strongest experimental constraint in their own paper. This mission asks: **(a)** what are the plausible *ways* 1a binds the Girdin C‑terminal sequence; **(b)** what is the more‑correct methodology; **(c)** does any candidate site *explain the SAR* (which signals, where, with what confidence, and what is explicitly not reproduced); **(d)** what happens when the authors' own pocket is put to the SAR test; **(e)** what evidence should drive the Stage‑2 go/no‑go on the two deferred items (GBA/Gαi3‑interface docking; phospho‑site / extreme C‑terminus).

---

## 2. (a) Candidate ways 1a binds — site hypotheses, confidence, and the FINAL lead/negative

The blind whole‑surface docking of the full six‑complex series across the compact ensemble members (Methodology Stages 3–4; `stage1/dock_blind/`, `stage1/site/`) clustered all poses into **four candidate sites** along the canonical sequence:

| site | canonical span | occupancy | complete 6‑complex SAR? | site MDE (kcal/mol) | clears both signals? | notes |
|---|---|---|---|---:|:--:|---|
| **B01** | 1692–1744 | 8/8 members, 286 poses | yes | 1.13 | **no** (A gap 0.32<MDE, B gap 0.05<MDE) | most‑occupied; overlaps GBA/Gαi3 + recurrent S01/S06 |
| B02 | 1752–1792 | 6 members | yes | 1.07 | **no** (A gap 0.58<MDE) | over Q1778 / S07 region |
| B03 | 1804 | 1 member | **no** (incomplete) | — | n/a | single‑occupancy, not scored |
| B04 | 1811–1833 | 6 members | yes | 0.72 | **no** (A gap 0.46<MDE) | **ρ=1.0/τ=1.0 but inside noise — NOT a lead** |

Source: `stage1/site/sar_by_site.json`, `stage1/site/blind_summary.json`.

**No site reproduces both robust SAR signals above its noise floor.** Blind verdict: **NEGATIVE — lead = none; expected false positives = 0.0** over all scored sites (720‑permutation null; `sar_by_site.json:expected_false_positives_sum_p = 0.0`). The `one_of_n_sites` list is **empty** — the search produced no "1 of N" candidate to promote.

**The anti‑gaming save (must not be undone).** Site **B04 (1811–1833)** has a *perfect* rank order — Spearman ρ = **1.0**, Kendall τ = **1.0** — yet is **correctly not a lead**: its Signal‑A gap (0.4627 kcal/mol) sits **below** the site MDE (0.7236), i.e. exactly the ~0.6 kcal/mol fine ordering that §0 declares unclaimable. A perfect ρ *inside the noise band* is the textbook multiple‑comparison false positive the guard exists to reject. **This report does not resurrect B04 as a hit.**

**FINAL lead/negative synthesis after rescoring (the decision this report is required to make).**
- The blind stage produced **no** SAR‑validated lead and **no** demoted "1 of N" site. There is therefore nothing that rescoring could *elevate* to a lead.
- Physics rescoring (GFN2‑xTB interaction energy) was run on the **best‑occupied** site **B01** — labeled as best‑occupied, not a validated lead (`stage1/rescore/rescore_results.json`). Result: **DOES NOT CONFIRM.** Rescoring MDE = **14.22 kcal/mol**; Signal A gap **−6.54** and Signal B gap **−3.97** are both far inside the band (R=H is not even least‑tight by mean xTB energy). The rescore mean‑rank anti‑correlates with Kd (ρ = −0.94, p = 0.005) but that is **inside the 14 kcal/mol noise band → not claimable in either direction**, and is reported for completeness only, *not* as "refuting" the SAR.
- **Therefore the FINAL determination is NEGATIVE: there is no SAR‑validated binding site.** No BLIND‑002 site is elevated (none qualified); rescoring corroborates nothing above its band, so no site is promoted on rescoring grounds either.

**The honest hypothesis space (confidence attached).**
- **H1 — shallow interaction near the GBA/Gαi3 helix (B01, 1692–1744). Confidence: LOW (structural hypothesis, not a validated lead).** B01 is the most‑occupied cluster (all 8 docked members). Its best‑occupied focus pose (**1d in E008, seed 101, affinity −9.27 kcal/mol**; `stage1/site/structural_rationalization.json`) places the 4‑R substituent in measured **3.32 Å** contact (PHE1743) — the R group is **not solvent‑exposed** (R_min 3.32 Å; R contacts PHE1743 3.32 Å, ILE1698 3.91 Å) — so an "R drives affinity" rationale is *geometrically admissible* here. B01 **overlaps the GBA/Gαi3 interface (canonical 1698–1702)** and recurrent transient pockets S01/S06. But because B01 **fails both MDE‑gated signals**, *no mechanistic story is asserted*; it is carried forward as the leading structural hypothesis to test, not as an answer.
- **H2 — a shape/size (non‑specific) binding mode.** The only place the coarse SAR reproduces (the authors' homology pocket, §5) reproduces it because R=H is a smaller ligand and amine rings are bulkier — effects that would emerge in almost any concave surface. This is the null‑ish hypothesis: 1a's affinity dependence on R may be **largely a property of the ligand's shape/size**, only weakly encoding a specific pocket. **Confidence: this shape/size contribution is real (high confidence it exists); that it pinpoints a physiological pocket — low.**
- **Rejected as a lead — B04 (1811–1833):** perfect rank order but inside noise (see above).

The candidate ways are thus: a **shallow, transient contact near the GBA/Gαi3 helix** (best structural hypothesis, low confidence), against a background in which the observed R‑dependence is at least partly **non‑specific shape/size**. **No deep, stable, SAR‑reproducing druggable pocket exists in the sampled ensemble** (288 fpocket pockets over 28 members → 8 recurrent sites, all *transient*: druggability spikes ~0.92 in single conformers but recurrence‑weighted mean low; `stage1/ensemble/pocket_catalogue.*`).

---

## 3. (b) The methodology (summary; full spec in METHODOLOGY.md)

The corrected protocol — full, re‑runnable, third‑party‑reproducible specification in [`METHODOLOGY.md`](METHODOLOGY.md) — is a seven‑stage pipeline:

**Stage 0** build & chemically validate the exact 1a–1e + R=H series (only 4‑R differs; +1 singlet octahedra; formulas < 0.7 mDa vs SI intact-cation HRMS for 4/5 — 1c has fragment-only SI HRMS and is EA/formula/graph-validated; `stage1/ligands_series/`) → **Stage 1** disorder‑aware ensemble of Girdin‑CT (28 distinct members / 5 independent source families; median Cα‑RMSD 20.3 Å, Rg 13–46 Å, pLDDT 23–42 — disorder represented honestly; `stage1/ensemble/`) → **Stage 2** per‑conformer fpocket transient‑pocket catalogue → **Stage 3** blind whole‑surface docking of the full series into **8 representative compact members (of 19 compact; the two extended ab‑initio monomer folds E001/E002 are not box‑dockable, and the 8 actually docked are Boltz‑co‑fold/homology‑derived — so the NEGATIVE rests on that biased subset; see METHODOLOGY.md Stage 3 / Limitation 4)** (smina, exh 32, Ir→Fe shape surrogate, rigid, xTB‑GFN2 charges; box = full bounding box + 8 Å; `stage1/dock_blind/`) → **Stage 4** SAR rank‑order validation against the fixed Kd yardstick with a pre‑registered **MDE** noise floor, pre‑registered **Signal A/B**, and a **720‑permutation multiple‑comparison guard** (`stage1/site/`) → **Stage 5** GFN2‑xTB physics rescoring of the lead/best‑occupied site at the same claim granularity (`stage1/rescore/`) → **Stage 6** orthogonal‑data consistency (Gαi3/GBA pulldown, mutagenesis residues).

The core correction versus the authors: **never assume a fold** (ensemble, not one rigid homology model), **search the whole surface blindly** (not one predefined pocket), and **make every site stand or fall against the measured SAR** under a noise‑gated, multiple‑comparison‑guarded test — converting docking from an illustration into a **falsifiable hypothesis test**, for which a rigorous NEGATIVE is itself a valid result.

---

## 4. (c) The SAR‑explanation verdict — which signals reproduced, where, with what confidence, and what is NOT

| signal | reproduced at a **blind ensemble** site? | reproduced in the **authors' homology pocket** (control)? | confidence / caveat |
|---|:--:|:--:|---|
| **Signal A** (R essential; R=H least tight ≥ MDE) | **NO** — 0/3 scored sites (best site B02 gap 0.58 < MDE 1.07) | **YES** — best‑score gap **0.952**, mean **0.938** ≫ MDE **0.125** | reproduces in the control **only because R=H is a smaller ligand** — a shape/size effect, **not pocket‑specific** |
| **Signal B** (amine > bulky/ether) | **NO** — 0/3 scored sites | **YES** — gap **0.298 / 0.340** > MDE **0.125** | ditto — amine rings are bulkier; **not pocket‑specific** |
| **Fine ordering** (1a<1d<1e<1c<1b, ~0.6 kcal/mol) | **NOT reproduced / NOT claimed anywhere** | **NOT claimed** (docking misplaces 1e) | at/below docking noise; B04's ρ=1.0 is exactly this and is **not** a claim |
| overall rank correlation (context only) | best blind site B01 ρ=+0.77 (p=0.07), B04 ρ=1.0 (inside noise) | ρ = **+0.83** (p=0.042), τ = **+0.73** (p=0.056) | correlation alone is **insufficient** — a site is a lead only if it clears the MDE‑gated signals **and** the guard |

Sources: `stage1/dock_ctrl/ctrl_results.json` (control signals A/B, ρ/τ); `stage1/site/sar_by_site.json` (blind sites); `stage1/rescore/rescore_results.json` (physics).

**Verdict.** The two **robust SAR signals are reproduced — but only in the authors' reproduced homology pocket, and only as a shape/size (non‑pocket‑specific) effect.** They are **not reproduced at any site the blind whole‑surface search actually nominates**, and physics rescoring of the best‑occupied blind site **does not confirm** them. **The fine ordering is explicitly not reproduced and is never claimed.** Net: the coarse dependence of Kd on R is a **real chemical effect that the method sees** (high confidence), but it is **not evidence of a specific physiological binding pocket** on this disordered region (the site question is answered NEGATIVE / hypothesis‑only).

---

## 5. (d) Authors'‑pocket‑control outcome

All six complexes (R=H mapped to the *actually‑measured* **Irppy‑bipy**, Kd > 20 µM) were docked into the **reproduced authors' homology "SH2" pocket** (`homology/girdin_sh2_model.pdb`, ProMod3/SWISS‑MODEL on mouse SOCS3‑SH2 2HMH), 6 seeds each, under the identical fixed protocol (`stage1/dock_ctrl/`).

**Result:** the pocket **reproduces BOTH** Signal A (best‑score gap **0.952** / mean **0.938** ≫ MDE **0.125**) and Signal B (gap **0.298 / 0.340** > MDE), with Spearman **ρ = +0.83 (p = 0.042)**, Kendall **τ = +0.73 (p = 0.056)** vs the primary‑source Kd (`ctrl_results.json`).

**Crucial honest reading (no overclaim).** Reproducing these coarse signals **does not validate the pocket**, for three converging reasons:
1. **The signals are not pocket‑specific.** Signal A is driven by R=H simply being a *smaller* ligand; Signal B by amine rings being *bulkier* — shape/size complementarity that would emerge in almost any concave surface. That is exactly why the **blind whole‑surface search + multiple‑comparison guard** (not one hand‑picked pocket) is the load‑bearing test — and *there* the same signals fail.
2. **The fold is refuted.** Prior work in this workspace established by **six independent methods** (InterPro/Pfam, HMMER PF00017 0 hits, TM‑align 0.11–0.13, Foldseek PDB+AFDB 0 hits, 19.6% identity) **plus a functional test** (SH2‑segment + EGFR‑pTyr co‑fold fails **0/5**, domain pLDDT 37–39, while the same method recovers a *bona fide* SH2 at 2.4 Å on a positive control) that Girdin‑CT is **not** a spontaneously‑folding SH2 domain (`altrecept/verdicts.json` VAL_MODEL_001/003/004; `REPORT_template_assessment_2026-07-06.md`). fpocket druggability at the authors' nominal pocket is **0.018**, and a ligand‑sized cavity exists in only **1/11** models. The pocket is a **homology artifact**.
3. **The fine ordering is still not recovered** even here (docking misplaces 1e).

**So the authors' pocket "explains" the coarse SAR only in the trivial shape sense, on a fold the target does not adopt** — which is why a single‑complex pose into it (as in the paper) is not falsifiable and does not constitute validation.

---

## 6. (e) Stage‑2 decision framing — evidence to drive the go/no‑go on the two deferred items

Per the mission charter, the two Stage‑2 items are **deferred and conditional** (user directive: *"do 2 and 3, then present results; decision on 1 and 4 depends on them"*). This section presents the Stage‑1 evidence that should drive that decision; it does **not** decide (out of worker scope).

### Item 1 — GBA/Gαi3‑interface targeted docking

**Evidence FOR a go:**
- The **GBA–Gαi3 interaction is the one experimentally‑validated interaction of this region.** The Girdin GBA motif (canonical ≈1672–1702) binds Gαi3, **reproduced** here vs PDB **6MHF** (`altrecept/verdicts.json` VAL_MODEL_002: REPRODUCED, 3/5 seeds pass interface_TM ≥ 0.5 & peptide backbone RMSD ≤ 4.0 Å; per‑seed interface_TM up to 0.564, RMSD 1.8–2.97 Å).
- **1a functionally disrupts the Girdin–Gαi3 interaction** (paper SI Figs S23/S24, grep‑verified present in `provenance/si/sapp.txt`) and **1a binds Gαi3 directly** (Kd ≈ 15 µM). A physiologically‑meaningful 1a site is therefore *expected to be consistent with* this interface.
- The blind search's **best‑occupied focus site B01 already overlaps the GBA/Gαi3 interface (1698–1702)** (`structural_rationalization.json`) — i.e. the disorder‑aware search independently drifted 1a toward this functional patch.

**Evidence AGAINST / for caution:**
- The GBA is a **thin surface helix with no deep pocket** (transient fpocket druggability only). A classical pocket‑docking pass may find only a shallow, low‑enrichment interface.
- The whole‑surface blind search **already covered this region (B01) and it did not reproduce the SAR above noise** — so a targeted pass should not be expected to convert it into a SAR‑validated lead; its value is **mechanistic** (does 1a compete with Gαi3?), not SAR‑explanatory.

**Go/no‑go driver:** Item 1 is the **higher‑value** Stage‑2 escalation — it targets the single validated function and a demonstrated 1a effect (Gαi3 disruption). A go is best framed as a **Gαi3‑competition experiment** (dock 1a at the GBA helix ± the Gαi3 partner), with realistic expectation of a shallow interface and *not* a fresh SAR‑reproducing pocket.

### Item 4 — phospho‑site / extreme C‑terminus (pS1717 / pY‑type sites; true C‑terminus 1815–1871)

**Evidence FOR a go:**
- The **extreme C‑terminus (canonical 1815–1871) was NOT sampled** in Stage 1 (modeling window 1690–1850) — a **genuine blind spot**. The authors' construct (1660–1870) includes this disordered tail, but this ensemble under‑covers it.
- **B04 (1811–1833), at the edge of that tail, had a perfect (but non‑claimable, inside‑noise) rank order** — a weak hint that the C‑terminal edge deserves a dedicated look (explicitly *not* a claim; §2).
- Phosphorylation was only *partially* represented (Boltz co‑folds carried SEP1717/PTR1744); a **phospho‑aware ensemble** could test whether a phospho‑group creates the basic‑amine‑recognition contact the SAR would need (Signal B favors basic amines; a phosphate is an obvious counter‑charge partner).
- Recurrent transient pockets **S01 (1743–1767, over R1746/K1750)** and **S07 (1778–1780, over Q1778)** cluster over basic/mutagenesis residues (`pocket_catalogue.*`) — plausible phospho‑gated sites.

**Evidence AGAINST / for caution:**
- **No stable druggable pocket exists anywhere** in Stage 1 (all 8 recurrent sites transient) — the region is a genuine IDR, so item 4 is inherently **exploratory**.
- Adding phospho‑forms and the extreme tail multiplies conformational sampling cost without a guarantee of a SAR‑reproducing site.

**Go/no‑go driver:** Item 4's strongest justification is the **unsampled 1815–1871 tail** (a real coverage gap) plus the **untested phospho‑gated hypothesis**; both are exploratory. It is the **lower‑confidence** of the two escalations but closes a genuine Stage‑1 blind spot.

**Overall Stage‑2 recommendation framing (for the user's decision, not a decision):** the Stage‑1 result is an honest NEGATIVE with a single low‑confidence structural hypothesis (B01/GBA). If Stage 2 is approved, **Item 1 (GBA/Gαi3 competition) is the higher‑value, better‑grounded escalation**; **Item 4 (phospho / extreme C‑terminus) closes the real coverage gap** but is more speculative. Neither is expected, on Stage‑1 evidence, to convert the NEGATIVE into a proven site; both would sharpen or falsify the H1 (GBA) / H2 (shape‑only) hypotheses. FEP/absolute‑ΔG remains out of scope unless a single site first cleanly reproduces rank order.

---

## 7. Confidence & limitations (honest boundaries)

1. **A binding site here is a computational hypothesis, never a proof.** The blind search returned a NEGATIVE; physics rescoring does not confirm; the authors' pocket reproduces only non‑specific shape/size signals. None of this proves *where* 1a binds — it constrains and falsifies.
2. **The fine ordering is not resolved, by design** (§0). Only Signals A and B are testable; B04's ρ=1.0 is inside noise and is not a claim.
3. **The MDE is conservative on this ensemble** (whole‑surface blind search over a heterogeneous IDR → large within‑complex variance → docking MDE 0.7–1.1, xTB MDE ~14 kcal/mol, both larger than the entire ~0.8 kcal/mol SAR spread). The test cannot certify a SAR sitting inside its own noise — it yields honest negatives but is biased toward false negatives by construction; a true weak signal could be missed.
4. **Ensemble sampling is incomplete and partly substituted.** The two GPU/host samplers named in the plan were environment‑blocked (vast.ai ColabFold AF‑subsampling blocked by the external‑code guardrail; OpenMM MD on hive t06 blocked by a saturating unrelated job) → **ANM normal‑mode expansion** was used as a labeled flexibility surrogate (not claimed to be MD). The extreme C‑terminus 1815–1871 is unsampled (see Item 4).
5. **Docking‑score / rigid‑body / gas‑phase limits.** smina/Vina has no Coulomb term; Ir is a Fe shape surrogate; the complex is rigid (justified for an inert octahedron; forbids induced fit). GFN2‑xTB rescoring is gas‑phase on a truncated, neutral, valence‑capped pocket — applied identically across complexes (so it does not bias the between‑complex comparison) but not a free‑energy calculation.
6. **Numbering discipline is load‑bearing** (paper = canonical − 1); a numbering error would silently move every contact and overlap.

---

## 8. Load‑bearing claims → evidence (handoff for independent re‑derivation)

A validator can re‑derive any three of the following to confirm nothing exceeds the evidence + oracle.

| # | Load‑bearing claim | Evidence path | Oracle / primary source |
|---|---|---|---|
| R1 | Experimental Kd SAR (1a 1.3 … R=H > 20 µM; rank 1a<1d<1e<1c<1b≪R=H) | `stage1/dock_ctrl/kd_primary_source.json` | `provenance/main_plain.txt` Fig 7E/8E/7G + SI S20 (quotes grep‑verified) |
| R2 | Blind search NEGATIVE: 4 sites, **0 clear both signals**, expected‑FP = 0.0; lead = none; `one_of_n_sites` empty | `stage1/site/{sar_by_site.json,blind_summary.json}` + `stage1/dock_blind/{pose_scores.csv,poses/}` | 720‑perm null + Kd oracle |
| R3 | **B04 ρ=1.0/τ=1.0 correctly NOT a lead** (Signal‑A gap 0.4627 < site MDE 0.7236 = fine ordering inside noise) | `stage1/site/sar_by_site.json` (site B04 block) | §0 unclaimable fine ordering |
| R4 | Authors' pocket reproduces BOTH Signal A (gap 0.952/0.938 ≫ MDE 0.125) & B (0.298/0.340) — ρ=+0.83 — but **non‑pocket‑specific** | `stage1/dock_ctrl/ctrl_results.json` | Kd oracle; `homology/girdin_sh2_model.pdb` |
| R5 | Physics rescore of B01 **DOES NOT CONFIRM**: MDE 14.22; Signal A −6.54 / B −3.97 inside band; ρ=−0.94 not claimable | `stage1/rescore/{rescore_results.json,README.md,verify_rescore.py}` | GFN2‑xTB re‑derive \|Δ\|=0.0000 |
| R6 | Best‑occupied focus pose 1d/E008 (−9.27): 4‑R in 3.32 Å contact (PHE1743), not solvent‑exposed; overlaps GBA/Gαi3 1698–1702; no pEGFR‑mutagenesis/acidic contact | `stage1/site/structural_rationalization.json` | Gαi3 6MHF (VAL_MODEL_002); SI S23/S24 |
| R7 | Series differs **only** in 4‑R; +1 singlet octahedra; formulas < 0.7 mDa vs SI intact-cation HRMS (4/5; 1c EA/formula) | `stage1/ligands_series/{graph_diff_check.json,SERIES_TABLE.md}` | SI HRMS/EA/NMR `provenance/si/sapp.txt` |
| R8 | Disorder‑aware ensemble: 28 distinct / 5 families; median 20.33 Å, min 2.06 Å; Rg 13–46 Å, pLDDT 23–42; 288 pockets / 8 transient recurrent sites | `stage1/ensemble/{diversity_summary.json,member_table.*,pocket_catalogue.*}` | UniProt Q3V6T2 Disordered 1736–1871 |
| R9 | Fold refuted (6 methods + 0/5 pTyr co‑fold; druggability 0.018; cavity 1/11) → authors' pocket is an artifact | `altrecept/verdicts.json` (VAL_MODEL_001/003/004), `REPORT_template_assessment_2026-07-06.md` | InterPro/HMMER/TM‑align/Foldseek + functional test |
| R10 | GBA–Gαi3 interaction reproduced (item‑1 driver): VAL_MODEL_002 REPRODUCED vs 6MHF (3/5 seeds) | `altrecept/verdicts.json` (VAL_MODEL_002) | PDB 6MHF; SI S23/S24 (1a disrupts pulldown) |
| R11 | Fixed docking protocol reproduces the 1a baseline within ±1.0 (Δ0.955); MDE defined once & imported verbatim downstream | `stage1/dock_setup/{PROTOCOL.md,MDE_ESTIMATOR.md,mde_estimator.py,verify_dock_setup.py}` | same‑engine smina reference −7.40 |

**Cited Stage‑1 outputs (all):** `stage1/ligands_series/` (LIG), `stage1/ensemble/` (ENS), `stage1/dock_setup/` (DOCK), `stage1/dock_ctrl/` (CTRL), `stage1/dock_blind/` + `stage1/site/` (BLIND), `stage1/rescore/` (RESCORE), `stage1/report/METHODOLOGY.md` (METH).

**Authorship of the outcome, stated plainly:** this mission does **not** claim to have found where 1a binds. Its deliverable is a **correct, reproducible, falsifiable protocol** and an **honest binding‑mode hypothesis space** — a rigorous NEGATIVE for the blind lead, a coarse‑signal‑only (non‑specific) result for the authors' refuted pocket, and a best‑occupied focus site (B01, GBA/Gαi3‑overlapping) carried forward as a **low‑confidence hypothesis, not a proof.**

---

## 9. Русское резюме (Russian summary)

### Вопрос миссии
Руан и соавт. (PNAS 2024;121(25):e2316615121) утверждают, что инертный октаэдрический Ir(III)-карбеновый комплекс **1a** связывается с C-концом человеческого белка Girdin/CCDC88A (UniProt Q3V6T2; конструкт His-Girdin-CT 1660–1870, Kd 1.3 µM по MST) в одном жёстком гомологическом «SH2-подобном» кармане — и **никогда не проверяли эту позу против измеренной серии Kd** (SAR), сильнейшего экспериментального ограничения в их же статье. Задача: найти правдоподобные **способы связывания 1a** и предложить **более корректную методику поиска сайта**, объясняющую зависимость Kd от структуры комплекса.

### Нумерация и «линейка» (только для чтения)
Все остатки — в канонической нумерации Q3V6T2; **в статье нумерация = каноническая − 1**. Экспериментальный ряд Kd (извлечён дословно из первоисточника `provenance/main_plain.txt`, Рис. 7E/8E/7G + SI Рис. S20): **1a 1.3 < 1d 1.5 < 1e 1.6 < 1c 2.6 < 1b 3.6 ≪ R=H (Irppy-bipy) > 20 µM**. Различается **только заместитель 4-R**. Робастных сигнала два: **A** — R существен (штраф R=H ≥ ~1.6 ккал/моль); **B** — грубый тренд «маленький основной амин > объёмный/эфир/трет-бутил». **Тонкий порядок среди аминов (~0.6 ккал/моль) — на уровне/ниже шума докинга и НЕ заявляется.**

### (a) Способы связывания 1a — гипотезы, уверенность и ИТОГ
Слепой докинг всей серии по всей поверхности ансамбля дал **4 сайта-кандидата** (B01 1692–1744; B02 1752–1792; B03 1804 — неполный; B04 1811–1833). **Ни один не воспроизводит оба робастных сигнала выше собственного шума.** Вердикт слепого поиска — **ОТРИЦАТЕЛЬНЫЙ**: лид отсутствует, ожидаемое число ложных срабатываний ≈ 0 (720 перестановок). Сайт **B04 имеет идеальный ранговый порядок (ρ=1.0)**, но **корректно НЕ является лидом**: его разрыв по сигналу A (0.46) ниже MDE сайта (0.72) — это как раз «тонкий порядок в шуме»; воскрешать его нельзя.

**ИТОГОВЫЙ вывод после переоценки (rescoring):** так как слепой этап не дал ни лида, ни отложенного сайта «1 из N», повышать нечего; физическая переоценка (GFN2-xTB) наиболее заселённого сайта **B01** даёт вердикт **«НЕ ПОДТВЕРЖДАЕТ»** (MDE = 14.22 ккал/моль; сигнал A −6.54, B −3.97 — внутри шума; антикорреляция ρ=−0.94 внутри шума → не заявляется). **Итог — ОТРИЦАТЕЛЬНЫЙ: SAR-валидированного сайта нет.** Единственная переносимая **структурная гипотеза (низкая уверенность)** — неглубокий переходный контакт у спирали **GBA/Gαi3 (B01, 1692–1744)**: лучшая поза (1d в E008, −9.27 ккал/моль) ставит 4-R в контакт 3.32 Å (PHE1743), группа R не экспонирована в растворитель, сайт перекрывает интерфейс Gαi3 (1698–1702) — но так как сигналы не проходят, механизм не утверждается. Стабильного глубокого «карманного» сайта в ансамбле нет (288 карманов fpocket → 8 переходных сайтов).

### (b) Методика
Семиэтапный воспроизводимый конвейер (полная спецификация — [`METHODOLOGY.md`](METHODOLOGY.md)): построение и валидация точной серии лигандов → **ансамбль с учётом разупорядоченности** (28 различных моделей / 5 источников) → каталог переходных карманов (fpocket) → **слепой докинг всей серии по всей поверхности** → **валидация ранга против фиксированной Kd** (MDE, сигналы A/B, гвардия множественных сравнений на 720 перестановках) → **физическая переоценка GFN2-xTB** → сверка с ортогональными данными. Ключевое отличие от авторов: **не постулировать фолд**, **искать по всей поверхности вслепую** и **проверять каждый сайт против измеренной SAR** — докинг становится **фальсифицируемым тестом**, а строгий отрицательный результат — валидным итогом.

### (c) Вердикт об объяснении SAR
Робастные сигналы **A и B воспроизводятся — но ТОЛЬКО в воспроизведённом гомологическом кармане авторов, и лишь как эффект формы/размера (не специфичный к карману)**. **Ни на одном сайте слепого поиска они не воспроизводятся**, а физическая переоценка B01 их **не подтверждает**. **Тонкий порядок нигде не воспроизведён и не заявляется.** Итог: зависимость Kd от R — реальный химический эффект, который метод «видит» (высокая уверенность), но это **не доказательство специфического физиологического кармана** (вопрос о сайте — ОТРИЦАТЕЛЬНЫЙ/гипотеза).

### (d) Контроль на кармане авторов
Все 6 комплексов, состыкованные в воспроизведённый гомологический «SH2»-карман (шаблон мышиного SOCS3-SH2, 2HMH), **воспроизводят оба сигнала** (A: разрыв 0.952/0.938 ≫ MDE 0.125; B: 0.298/0.340; ρ=+0.83, p=0.042). **Но это не валидирует карман:** сигналы обусловлены формой/размером (R=H меньше, амины объёмнее) и **не специфичны к карману**; сам фолд **опровергнут** 6 методами + функциональным тестом (co-fold с pTyr EGFR: 0/5; друггабильность 0.018; полость лишь в 1/11 моделей). То есть карман «объясняет» грубую SAR лишь тривиально по форме, на фолде, который мишень не принимает.

### (e) Рамка решения для Этапа 2 (пункты 1 и 4 — отложены, решает пользователь)
- **Пункт 1 (докинг к интерфейсу GBA/Gαi3):** ЗА — это **единственное экспериментально подтверждённое взаимодействие региона** (VAL_MODEL_002 воспроизведён против PDB 6MHF, 3/5 сидов); 1a разрушает связь Girdin–Gαi3 (SI S23/S24) и связывает Gαi3 напрямую (Kd ≈ 15 µM); слепой поиск сам сместил 1a к этому интерфейсу (B01 перекрывает 1698–1702). ПРОТИВ — GBA это тонкая поверхностная спираль **без глубокого кармана**; слепой поиск уже покрыл этот регион и SAR там не прошла. **Драйвер:** более высокоценная эскалация — как эксперимент **конкуренции с Gαi3**, с ожиданием мелкого интерфейса, а не нового SAR-кармана.
- **Пункт 4 (фосфосайт / крайний C-конец 1815–1871):** ЗА — крайний C-конец **не был просэмплирован** на Этапе 1 (окно 1690–1850) = реальный пробел; B04 (1811–1833, на краю хвоста) дал идеальный, но «шумовой», ранг; фосфо-формы представлены лишь частично — фосфат мог бы создать недостающий контакт для основных аминов; переходные карманы S01 (над R1746/K1750) и S07 (над Q1778). ПРОТИВ — стабильного кармана нет нигде, регион разупорядочен → чисто разведочно. **Драйвер:** закрывает реальный пробел покрытия, но более спекулятивен.
- **Общая рамка (для решения пользователя, не решение):** результат Этапа 1 — честный ОТРИЦАТЕЛЬНЫЙ с одной низкоуверенной гипотезой (B01/GBA). Если Этап 2 одобрен — **Пункт 1 обоснован лучше**; **Пункт 4 закрывает пробел покрытия**, но спекулятивнее. Ни один по данным Этапа 1 не обещает превратить отрицательный итог в доказанный сайт; FEP/абсолютная ΔG — вне области, пока какой-либо сайт чисто не воспроизведёт ранг.

### Ограничения (кратко)
Сайт — вычислительная **гипотеза, не доказательство**; тонкий порядок не разрешается по построению; MDE консервативен (склонен к ложноотрицательным); сэмплирование ансамбля неполное (AF-субсэмплинг/MD были заблокированы средой → ANM как помеченный суррогат; хвост 1815–1871 не покрыт); докинг — жёсткое тело, Ir→Fe (форма), без кулоновского члена; xTB — газовая фаза на усечённом нейтральном кармане (идентично для всех комплексов, поэтому не смещает сравнение между комплексами). Дисциплина нумерации (статья = каноническая − 1) — несущее допущение.

**Каждое несущее утверждение прослеживается к файлу вычисленных доказательств + первоисточнику Kd (таблица в §8); переоценки/переоформления результатов не производилось; тонкий порядок и «доказанный сайт» не заявляются.**