Figure 1 - Delta Vina score vs WT v5 for each mutant (apo blue, holo orange).
Grey bars = mis-docked (RMSD > 3 A) or low-confidence (n_modes < 5).
Grey band = Vina noise floor +/-0.85 kcal/mol. Positive = destabilising.
The v5 fix. Round-4 reviewers identified that v4 placed the cofactor by Kabsch-aligning the CCD-ideal D16 onto the bound conformer, yielding a 2.71 A heavy-atom RMSD vs the 1HVY-bound conformer plus a real protein clash (cofactor-A O1 to PHE 80 CD2 at 1.95 A). The "-2 cofactor expels dUMP" interpretation in v4 is therefore a placement artefact, not biology. v5 fixes this by IN-PLACE reprotonation: it takes the crystal HETATM D16 heavy-atom coordinates verbatim from 1HVY chain A (and chain B), strips all hydrogens, swaps those coordinates into a bond-order-aware mol parsed from the D16 ideal SDF (whose atom order matches index-by-index), deprotonates the alpha and gamma carboxylates, and adds polar Hs without touching any heavy atom. A hard assertion would abort if any heavy atom moved by > 0.001 A. A clash gate aborts if any cofactor heavy atom is within 1.8 A of a protein heavy atom.
What changed in WT-holo. Top affinity moved from -5.24 kcal/mol (v4, mis-docked at 12.95 A from crystal) to -8.25 kcal/mol (v5, RMSD 0.33 A from crystal). This is now within ~1 kcal/mol of the apo result (-9.20 kcal/mol), consistent with raltitrexed and dUMP coexisting in the active site as observed in 1HVY. The v4 "expulsion" effect is gone.
What did NOT change in v5. Even with the cofactor placed correctly, no individual mutant exceeds Vina's noise floor on holo. The top destabiliser (R215A_N226A double, +0.77 kcal/mol) is still below +/-0.85 kcal/mol. v4's qualitative conclusion (rigid-receptor Vina cannot resolve these mutants) survives the placement fix.
| Condition | Top affinity (kcal/mol) | mean top-3 | n modes | RMSD top-pose vs crystal dUMP (A) | Best seed | Affinity range across seeds (kcal/mol) | Source |
|---|---|---|---|---|---|---|---|
| WT apo | -9.20 | -8.78 | 27 | 0.91 | 42 | 0.12 | reused from v4 (apo receptor unchanged) |
| WT holo | -8.25 | -7.15 | 2 | 0.33 | 13 | 0.04 | v5 (in-place reprotonated cofactor) |
v5 WT-holo selection: lowest top affinity across seeds {42, 7, 13, 99, 256} at exh=96; if max n_modes < 10, fallback {1, 2025, 31337} at exh=128. The v5 WT-holo n_modes is low because the binding funnel collapses to a single dominant pose (RMSD ~0.3 A) once the cofactor is correctly placed, with little alternative. This is consistent with high-confidence binding, not poor sampling.
Panel: 20 mutants - 8 ala-scan, 7 opposite-charge, 5 doubles, 4 arg-clamp, 1 surface control (T170A); G217W dropped per v3 (helix-break). Each mutant docked under apo and holo conditions. Holo dockings rebuilt in v5 against the in-place reprotonated cofactor receptor; apo dockings reused from v3 (apo receptor unchanged across v3->v4->v5).
The C195A row, when listed, is highlighted in pink AND its mis_docked/low_confidence flags are shown explicitly (sci-off review item 4).
| Mutant | category | Delta Vina | RMSD (A) | n modes | flags |
|---|---|---|---|---|---|
| R215E | opposite | +0.61 | 2.43 | 19 | ok |
| R50A | ala_scan | +0.58 | 2.42 | 20 | HOLO_mis_docked |
| R215A | ala_scan | +0.57 | 2.25 | 20 | ok |
| Q214A | ala_scan | +0.54 | 2.36 | 19 | HOLO_low_conf |
| N226D | opposite | +0.42 | 2.40 | 20 | HOLO_low_conf |
| Mutant | category | Delta Vina | RMSD (A) | n modes | flags |
|---|---|---|---|---|---|
| R215A_N226A | double_substrate_orient | +0.77 | 0.60 | 9 | APO_mis_docked |
| H196A | ala_scan | +0.49 | 0.34 | 5 | ok |
| R215E | opposite | +0.47 | 0.39 | 5 | ok |
| R215A | ala_scan | +0.44 | 0.40 | 5 | ok |
| Metric | Value |
|---|---|
| Pearson r (apo vs holo Delta, well-docked, n_modes>=5) | -0.740 (n=3), p = 0.470 |
| Spearman rho (apo vs holo, same set) | -0.500, p = 0.667 |
| Conclusion | No statistically significant apo-holo correlation (both p > 0.19; sci-off review item 3) |
| Vina noise floor (kcal/mol) | +/-0.85 (Trott & Olson 2010; Forli et al. 2016) |
| n mutants with |Delta_holo| > noise floor | 0 |
Per sci-off review item 3, the filtered Spearman rho (|Delta| > 0.3) computed in v4 is omitted from this table because in v5 it would have n < 5; the n=4 figure cited in v4 was statistically meaningless and would invite cherry-picking.
The v4 report attributed the WT-holo dock landing 12.95 A from the crystal pocket to "narrow funnel" and tentatively to "-2 ionised cofactor expelling -2 dUMP via electrostatics". Round-4 structural-bioinformatics review showed that this interpretation was wrong on two counts:
v5 confirms the diagnosis. With the cofactor heavy atoms placed at their crystal coordinates (RMSD 0.000 A vs 1HVY) and the carboxylates correctly deprotonated (canonical SMILES contains [O-] twice), and 0 protein clashes < 1.8 A, the WT-holo top pose lands in the canonical pocket (0.33 A from crystal dUMP, top affinity -8.25 kcal/mol). The v4 "expulsion" finding does not survive the placement correction.
| Stage | Tool / parameters |
|---|---|
| Cofactor reprotonation (v5) | IN-PLACE: HETATM D16 heavy-atom coords from 1HVY chain A and chain B, swapped into a bond-order-aware mol parsed from the D16 ideal SDF (atom order matches index-by-index); explicit -COOH -> COO- deprotonation; AddHs(addCoords=True). Hard assertion: no heavy atom moved > 0.001 A. Hard gate: no cofactor-protein heavy-atom pair < 1.8 A. v5 cofactor heavy-atom RMSD vs 1HVY = 0.000 A; protein clashes < 1.8 A = 0. |
| Receptor prep (Apo & Holo) | obabel -xr -p 7.4 --partialcharge gasteiger; max abs charge 0.507. |
| WT docking (apo) | Reused from v4: Vina 1.2.7, exh=96, num_modes=32, box=22^3 A, seeds {42, 7, 13, 99, 256}. |
| WT docking (holo, v5) | Vina 1.2.7, exh=96, num_modes=32, box=22^3 A, seeds {42, 7, 13, 99, 256}; fallback to {1, 2025, 31337} at exh=128 if max n_modes < 10 (fallback fired in v5; final selected best-seed used exh=96). |
| WT selection | lowest top_affinity (NOT RMSD); tie-break highest n_modes. |
| Mutant docking (holo, v5) | Vina 1.2.7, exh=32, num_modes=20, box=22^3 A, seed=42. |
| Mutant docking (apo) | Reused from v3 (apo receptor unchanged). NOTE protocol asymmetry below. |
| UMP atom-name preservation | walk PDBQT in heavy-atom order; transplant names from input ligand_h.pdb. |
| Sign convention | delta_vina_vs_wt = top_aff_mut - top_aff_wt_v5 (positive = destabilising). |
| mean_topk | mean(affinities[:min(3, n_modes)]). |
| mis_docked filter | RMSD top-pose vs crystal dUMP > 3 A. |
| low_confidence filter | n_modes < 5 (holo only). |
| Statistics | Pearson r and Spearman rho on the well-docked subset (n_modes>=5, RMSD<=3); filtered Spearman from v4 dropped (n=4 too small). |
In v4, with a placement-buggy cofactor, WT-holo itself landed RMSD 12.95 A from the crystal dUMP, so v4 reported holo using a relaxed metric (|RMSD - WT_holo_RMSD| > 3 A). In v5 the WT-holo RMSD is 0.334 A < 3 A, so the standard mis_docked = (RMSD > 3 A vs crystal) definition applies and the v4 dual-reference relaxation is not needed. This caveat is preserved here to make the methodological dependency explicit.
WT-apo and WT-holo were docked with a 5-seed sweep at exh=96 (and exh=128 fallback). Mutant-apo dockings are inherited verbatim from v3, which used a single seed at exh=32. Mutant-holo dockings (in v5) likewise use single seed exh=32. The Delta_apo column therefore mixes a high-effort WT reference with low-effort mutant runs - the apo-side mutant numbers are slightly noisier than the holo-side mutant numbers but the WT-relative reference is consistent within each condition.
AutoDock 4 / Vina use a united-atom convention: non-polar hydrogens are merged into their parent heavy atoms, and polar hydrogens (HD type) carry zero partial charge with the H-bond contribution folded into the parent. Maximum heavy-atom |q| in the v5 receptor is 0.507 (well above the 0.05 sanity threshold).
Outputs: 03e_structure_v5/, 06e_docking_wt_v5/, 07e_mut_docking_v5/, 08e_analysis_v5/, 09e_report_v5/. v1, v2, v3, v4 left untouched.