Tc-43/THRB_Agonist_Designs_1R6G

THRβ Agonist Designs — 1R6G (GA-II)

Why this target matters. Thyroid hormone drives lipid clearance and hepatic fat oxidation, but thyroid hormone itself is unusable as a drug because THRα engagement causes tachycardia, arrhythmia and bone loss. Resmetirom's 2024 approval for MASH proved that a β-selective thyromimetic can separate the metabolic benefit from the cardiac liability — and made β-over-α selectivity the whole game for this target.

188 small molecules generated de novo by the Technetium TC-43.ai engine (GA-II), conditioned on the ligand-binding pocket of the human thyroid hormone receptor β ligand-binding domain, using the 3.0 Å crystal structure 1R6G — THRβ LBD with a β-selective thyromimetic — 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.

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

---

Read this first: these designs do not engage the β-selectivity residue

THRβ selectivity has a single well-defined structural origin. Asn331 in THRβ is a serine (Ser277) in THRα — it is the one amino-acid difference in the ligand pocket, and it is the handle every β-selective thyromimetic is designed around.

No design in this set reaches it.

Measurement across all 188 designs	Result
Closest heavy atom to the Asn331 side chain (OD1/ND2)	median 6.31 Å (min 4.42, max 6.90)
Designs within H-bond range (< 3.6 Å) of Asn331 side chain	0 / 188
Designs within 4.5 Å	1 / 188

Asn331 appears at 100% in the contact table below, but that is backbone contact — the residue lines the pocket. Its side-chain amide, the part that distinguishes β from α, is never engaged.

Consequence: this set carries no structural basis for β-over-α selectivity. It should be read as THRβ-pocket binders, not as selective agonists, and a THRα counter-model is the obvious next step before any of it is taken further. This is published as a negative result worth having, not a hidden one — asn331_sidechain_dist is a column in designs.csv so you can check it yourself.

What the designs do engage

Residue	Role	Designs in contact (4.5 Å)	H-bond range (< 3.6 Å)
His435	canonical 4′-OH partner of T3 and thyromimetics	184 / 188 (98%)	108 / 188 — but carbon is the nearest atom in 67, so true H-bond geometry in ~32
Arg320	anion-binding cluster	185 / 188 (98%)	61 / 188, all via oxygen
Arg282	anion-binding cluster	18 / 188 (10%)	1 / 188
Arg316	anion-binding cluster	4 / 188 (2%)	0 / 188
Phe272, Ile276, Met313, Leu330, Leu346	hydrophobic core	188 / 188 (100%)	—
Met442, Phe455, Phe439, Ala317, Thr329	pocket wall	180–184 / 188 (96–98%)	—

So one arm of the anion-binding cluster (Arg320) is picked up by a carbonyl oxygen in a third of designs, and His435 is reached mostly as a hydrophobic contact. Median 20 contact residues per design (range 16–25).

Physicochemistry — a different class from every known thyromimetic

The designs are basic, where T3 and sobetirome are acidic and resmetirom is neutral. Median 3 basic nitrogens per molecule (range 2–6); only 3 of 188 carry any acidic group.

Properties below are for the neutral parent, which is the fair comparison — the source SMILES are written in a protonated form, and quoting cLogP on the cation would understate lipophilicity by roughly 2.5 log units.

Property	This set (neutral parent)	T3	Sobetirome (GC-1)	Resmetirom
MW	299 – 454 (median 340)	651	286	366
cLogP	−1.65 – 3.08 (median 0.86)	+3.95	+3.74	+4.30
TPSA	35.6 – 111.5 (median 66.9)	—	—	—
QED	0.49 – 0.94 (median 0.82)	—	—	—
Acidic group	3 / 188	carboxylate	carboxylate	none (cyanoazauracil)

QED at 0.82 median is high and molecular weight is modest. But the set sits about 3 log units more polar than every thyromimetic that has worked, and carries basic amines none of them have. Whether that is genuine novelty or generator drift into amine-rich space is exactly the question a reader should bring to it. Resmetirom does show the carboxylate is not mandatory — being acid-free is precedented; being polybasic is not.

Docking	Range	Median
Vina score	−12.30 to −11.00 kcal/mol	−11.30
Ligand efficiency	0.44 – 0.52	0.46
Heavy atoms	22 – 25	24

Maximum ECFP4 similarity of the neutral parents to the references: 0.197 (T3), 0.169 (sobetirome), 0.128 (resmetirom).

Further caveats

• This is one chemotype. All 188 designs contain a piperazine and a γ-lactam — 188/188 for both. The 149 distinct Murcko scaffolds and 176 unique SMILES count decorations around that constant pair. Do not read the scaffold count as chemotype diversity.
• Score-filtered. The −11.0 to −12.3 kcal/mol window is a selection artefact; vina_score will not rank these against each other.
• 3.0 Å receptor with high B-factors (structure mean B = 74). Side-chain positions, including Asn331's, are softly determined — weight backbone contacts accordingly. The Asn331 result above is a ~6 Å miss, well outside the range where that softness would change the conclusion.
• Rigid-receptor docking. Vina scores are not affinities, and Vina has no electrostatic term — relevant here, because a set this charged is exactly where that omission bites.
• Nothing here has been synthesised or assayed.

Contents

File	Description
designs.csv	One row per design: ID, SMILES (as generated) and neutral parent, Vina score, ligand efficiency, formal charge, basic-nitrogen count, neutral-parent properties, asn331_sidechain_dist, his435_sidechain_dist, arg320_guanidinium_dist, contact residues, Murcko scaffold
ligands.sdf	All 188 docked poses in 3D, bond orders and formal charges from the source SMILES; verified by stereo-stripped canonical round-trip (188/188) and charge preservation (188/188)
receptor_1R6G.pdb	The shared THRβ LBD receptor, reference thyromimetic absent — the frame all poses are in
structures/	Per-design ligand-only pose (<design_id>.pdb), 188 files, in the receptor frame

Merge receptor_1R6G.pdb with any structures/*.pdb to rebuild a complete complex.

Usage

import pandas as pd
df = pd.read_csv("designs.csv")

# the selectivity question, front and centre
print(df.asn331_sidechain_dist.describe())
print((df.asn331_sidechain_dist < 3.6).sum(), "designs H-bond the beta-selectivity residue")

# designs that at least reach the Arg320 arm of the anion cluster
print((df.arg320_guanidinium_dist < 3.6).sum(), "of", len(df))

Citation

Molecules generated by the Technetium TC-43.ai engine (GA-II) at Technetium Therapeutics. Receptor from PDB 1R6G. Poses scored with AutoDock Vina. Structural analysis and curation by Claude Code.