Tc-43/Nav1.7_VSD4_NonAnionic_Designs_8I5G

Nav1.7 VSD4 Non-Anionic Designs — 8I5G (GA-II)

Why this target matters. Human genetics validates Nav1.7 as thoroughly as any pain target in existence — loss-of-function carriers are congenitally insensitive to pain, gain-of-function mutations cause erythromelalgia — yet no Nav1.7 blocker has reached approval, and the clinical failures of the aryl sulfonamide class are widely attributed to the properties that come with its anionic warhead.

459 small molecules generated de novo by the Technetium TC-43.ai engine (GA-II), conditioned on the voltage-sensing domain IV (VSD4) of human Nav1.7, using the 2.7 Å cryo-EM structure 8I5G as the receptor. Each molecule was constructed against this pocket rather than selected from a compound library — docking (AutoDock Vina) came afterwards, to place and score the generated molecules in the site.

These designs are not anionic. That is the point of the set.

Target site definition, structural analysis, interaction profiling and dataset curation by Claude Code. Molecule generation by the Technetium TC-43.ai engine (GA-II).

---

The anionic problem, and what this set does instead

Every disclosed VSD4-targeted Nav1.7 inhibitor of the aryl sulfonamide class — PF-05089771, GX-936 and their relatives — carries an acidic aryl sulfonamide that is deprotonated at physiological pH. The anion pairs with the fourth gating arginine of the DIV-S4 helix, and that interaction is what buys the class its subtype selectivity. It also brings high plasma protein binding, poor CNS and nerve-tissue distribution, and pH-dependent behaviour — liabilities that track the charge rather than the scaffold.

This set was generated against the same pocket without that warhead:

	designs	share
Contain any sulfonamide	0	0.0%
Contain an aryl sulfonamide (the class warhead)	0	0.0%
Contain a tetrazole or acyl sulfonamide	0	0.0%
Not net-anionic at pH 7.4	443	96.5%
Net-anionic	16	3.5%

Charge distribution: 323 neutral, 119 cationic (+1, a protonated aliphatic amine), 16 anionic (−1), 1 dicationic. The 16 anions carry 7 carboxylates and 9 phenolates — not a single sulfonamide among them. Filter on is_anionic = 0 for the non-anionic subset.

Maximum ECFP4 Tanimoto similarity to the anionic references:

Reference	Max similarity across 459 designs
PF-05089771	0.133
GX-936	0.167

How a neutral ligand holds the arginine anchor

The interesting result is not that the anion is absent — it is that the anchor is still engaged.

All 459 designs contact R1619, the fourth DIV-S4 gating arginine, at a median closest-approach of 2.89 Å. In 423 of 459 the nearest ligand heavy atom is an oxygen; in 33 it is a nitrogen. The guanidinium is being held by a neutral hydrogen-bond acceptor where the aryl sulfonamides put a formal negative charge.

The second anchor is the S3 acidic face: 309/459 designs come within 4.0 Å of D1597, most often through a nitrogen or an N–H — the amino group of the core donating to the aspartate.

Contact frequency across the 459 poses (chain A, 4.5 Å heavy-atom cutoff, 8I5G numbering):

Residue	Element	Designs in contact
TYR1548	S2 aromatic	459 / 459 (100%)
ASN1551, VAL1552	S2	459 / 459 (100%)
ILE1588, SER1589	S3	459 / 459 (100%)
GLY1592, MET1593, ALA1596	S3	459 / 459 (100%)
ASP1597	S3 acidic	459 / 459 (100%)
GLU1600	S3 acidic	459 / 459 (100%)
PHE1609	S4	459 / 459 (100%)
ARG1613 (R2)	S4 gating charge	459 / 459 (100%)
ALA1615	S4	459 / 459 (100%)
ARG1616 (R3)	S4 gating charge	459 / 459 (100%)
ARG1619 (R4)	S4 gating charge	459 / 459 (100%)
ILE1555	S2	456 / 459 (99.3%)
TRP1549	S2 aromatic	433 / 459 (94.3%)
GLU1545	S2 acidic	311 / 459 (67.8%)
ILE1612	S4	248 / 459 (54.0%)
PHE1594	S3	150 / 459 (32.7%)

Median 19 contact residues per design (range 17–21). The binding mode is highly uniform.

Property profile

Property	Range	Median
Docking score (AutoDock Vina)	−16.20 to −14.00 kcal/mol	−14.30
Ligand efficiency	0.47 – 0.56	0.49
Heavy atoms	26 – 30	30
Molecular weight	351.4 – 468.3 Da	400.5
cLogP	0.19 – 4.00	2.32
TPSA	86.7 – 132.9 Ų	108.7
H-bond donors	1 – 3	1
H-bond acceptors	6 – 10	7
Rotatable bonds	3 – 8	5
Fsp³	0.20 – 0.50	0.29
QED	0.41 – 0.77	0.56

Ligand efficiency sits at 0.49 median on 30 heavy atoms, against PF-05089771's 500 Da and 31 heavy atoms — these are smaller, less lipophilic molecules than the class they are meant to replace.

Read this before using the set

• This is one chemotype, not 459 independent ideas. Every design contains the same 2-aminotriazin-4(3H)-one linked by a methylene to a quinoline. The 321 distinct Murcko scaffolds and 440 unique SMILES count decorations around that constant core — treat the set as a dense SAR fan around a single hypothesis, and do not read the scaffold count as chemotype diversity.
• Score-filtered. The narrow −14.0 to −16.2 kcal/mol window is a selection artefact, not a property of the generative run. vina_score will not rank these usefully against each other.
• The receptor conformation was induced by an anionic ligand. 8I5G was solved with PF-05089771 bound; that molecule is removed here, but the VSD4 side chains — including the R1619 rotamer — are the ones its charge selected for. Neutral ligands are being fitted to a pocket shape an anion produced. This is the central caveat of the set and the first thing to test with a flexible-receptor or MD follow-up.
• Rigid-receptor docking. Vina scores are not affinities.
• Nothing here has been synthesised or assayed. These are computational hypotheses.

Contents

File	Description
designs.csv	One row per design: ID, SMILES, Vina score, ligand efficiency, formal charge, is_anionic flag, acidic-group count, physicochemical properties, per-design contact residues, Murcko scaffold
ligands.sdf	All 459 docked poses in 3D, bond orders and formal charges assigned from the source SMILES; verified by stereo-stripped canonical round-trip (459/459) and charge preservation (459/459)
receptor_8I5G.pdb	The shared Nav1.7 receptor, PF-05089771 removed — the frame all poses are in
structures/	Per-design ligand-only pose (<design_id>.pdb), 459 files, in the receptor frame

Poses are shipped ligand-only against one shared receptor because the full complexes are ~1 MB each; merge receptor_8I5G.pdb with any structures/*.pdb to rebuild a complete complex.

Usage

import pandas as pd
from rdkit import Chem

df = pd.read_csv("designs.csv")

# the non-anionic subset — the reason this set exists
neutral = df[df.is_anionic == 0]
print(len(neutral), "of", len(df))

# designs that hold the R4 gating arginine (all of them, by construction of the site)
print(df.contact_residues.str.contains("ARG1619").mean())

poses = [m for m in Chem.SDMolSupplier("ligands.sdf") if m]

Citation

Molecules generated by the Technetium TC-43.ai engine (GA-II) at Technetium Therapeutics. Receptor from PDB 8I5G. Poses scored with AutoDock Vina. Structural analysis and curation by Claude Code.