Tc-43/GPR75_Inhibitor_Designs_GA-II

GPR75 Inhibitor Designs — Technetium GA-II

139 small molecules generated de novo by the Technetium `TC-43.ai` engine (GA-II), conditioned on the extracellular vestibule of GPR75, a first-in-class anti-obesity target. 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. Each is docked into a rigid receptor afterwards and supplied as a full protein–ligand complex.

GPR75 is unusual: its orthosteric pocket is sealed shut. This design set therefore targets a non-orthosteric vestibule site defined from the experimental cryo-EM structure.

Provenance

1. Drug target rationale

Summarised from the Apodex AI structure-based target dossier, 2026-08-31.

GPR75's strongest validation is human genetics. Sequencing of ~640,000 exomes identified protein-truncating loss-of-function variants in GPR75 in about 4 per 10,000 individuals. Heterozygous carriers show a lean, obesity-protected phenotype:

Because loss of function is protective, the rational therapeutic direction is inhibition or inverse agonism — not agonism. Human genetics also sets the efficacy ceiling: near-complete inhibition of GPR75 signalling is what the carrier phenotype emulates.

Expression and signalling

Broadly expressed across tissues relevant to energy balance: CNS including hypothalamus, adipose, pancreatic islets, vascular endothelium and smooth muscle, heart, kidney, retina.

The dominant pathway is Gα_q → PLC → IP₃ → Ca²⁺ → PKC. Downstream of Gα_q, GPR75 engages GIT1–EGFR–MAPK/NF-κB signalling in vascular and inflammatory contexts. Modest constitutive Gα_i activity is also reported — basal signalling without added ligand.

Endogenous ligands (debated)

GPR75 remains an IUPHAR orphan. 20-HETE is the leading candidate, binding with high apparent affinity and activating Gα_q/PLC/IP₃/Ca²⁺ signalling. CCL5 (RANTES) is proposed as a modulator, but β-arrestin recruitment and direct binding are inconsistent across systems.

Disease linkage

• —Obesity & metabolic disease — LOF carriers are obesity-protected; knockout mice are lean with better glucose tolerance and insulin sensitivity on high-fat diet.
• —Cardiovascular & renal — 20-HETE/GPR75 signalling promotes vascular smooth-muscle contraction, endothelial dysfunction and hypertension; blockade lowers blood pressure in models.
• —Liver / MASLD-NAFLD — expression and 20-HETE signalling correlate with steatosis and inflammation; inhibition attenuates liver fat preclinically.
• —Neuroinflammation & cancer — contributes to microglial activation; promotes proliferation, invasion and metastasis via EGFR/MAPK and NF-κB.

2. Why the vestibule, and not the orthosteric pocket

Cryo-EM changed the design problem. Structures of human GPR75 — 9XQC (Gq-coupled, 3.0 Å), 9XQN (apo, 3.91 Å) and an NbH3-stabilised active-like complex (~3.6 Å) — show a constitutively active-like receptor whose orthosteric pocket is occluded:

• —The canonical DRY motif is replaced by HRL, and the sodium pocket is collapsed, carrying a noncanonical Lys^7.45 that helps stabilise the active-like state.
• —ECL2 folds back into the orthosteric site via a CLPM motif, with M192 plugging a deep hydrophobic sub-pocket. A C118 (TM3)–C189 (ECL2) disulfide locks this cap in place, largely blocking solvent access. MD shows the capped state is stable.
• —The pocket that remains has a hydrophobic floor (C214^5.50, V330^6.44, C334^6.48) with a small hydrophilic wall patch (H122^3.29, S125^3.32, S126^3.33, E358^7.35).

Measured on the receptor models used here, the orthosteric channel radius is 0.84–1.4 Å against a 1.7 Å ligand heavy-atom requirement — the cavity is impassable in every state examined. Classical orthosteric docking is therefore not a viable route, and the dossier's recommendation is to pursue allosteric, ECL2-adjacent or lateral-entry strategies.

This set follows that recommendation and targets the extracellular vestibule.

3. The receptor and the site

Receptor

receptor_GPR75_inactive_model.pdb — an inactive-state transition model built from 9XQC by consensus per-helix active→inactive transition derived from five class A active/inactive pairs (β2AR, A2A, 5-HT2A, M2R, NTSR1), transferred via GPCRdb generic numbering, then restrained-minimised with 150 ps MD. CA RMSD versus 9XQC is 1.33 Å, zero clashes below 2.2 Å. Docking was rigid — chain R is byte-identical across all 139 complexes, so poses superpose without alignment.

Both experimental GPR75 structures are active-state; no experimental inactive structure exists.

Vestibule site definition

Volume 142.7 ų, enclosure 6.10/7. The site was validated by docking a 2-methylbenzamide probe independently into 9XQC and into the inactive model — the poses agree to 2.45 Å centroid with 57% shared contacts. 9XQN (3.91 Å) places the probe elsewhere and is treated as unreliable.

Eight residues are reproducible contacts across both structures — the validated core:

V104^2.63, L105^2.64, F107^2.66, A110^2.69, L188 (ECL2), C189 (ECL2), Y355^7.32, L359^7.36

4. The designs

Chemistry

Two chemotypes account for essentially the whole set:

Diversity sits in the exit vector rather than the core — 120 scaffolds over 139 designs.

Binding mode

Recomputed from the deposited complexes. Contact = any ligand heavy atom within 4.5 Å.

The set reproduces the validated vestibule core almost perfectly — mean 7.9 of the 8 reproducible residues, with 122/139 engaging all eight. All 139 poses form a single cluster (maximum pairwise centroid spread 5.0 Å).

Lys134 and Asp210 — the salt bridge that stabilises the active state — are present in the model and contacted by 0/139. This is not that site.

The anchor

128 of 139 donate a primary aryl amine N–H to the Leu105 backbone carbonyl, N···O mean 3.03 Å (range 2.59–3.20 Å). Every remaining pose donates some N–H to the same acceptor, and all 139 carry a primary aryl amine. That single hydrogen bond is the pharmacophore of the series.

Cross-check against the experimental structure

The site definition requires hits to score in both receptors. Because the inactive model was built onto 9XQC, the two share a coordinate frame, so poses transplant directly:

The series is sterically compatible with the active cryo-EM structure — 90% clash-free without any re-docking. The contact fingerprint degrades though, chiefly at H187 and S109, which the site definition already flags as inactive-model-specific and where vestibule side chains deviate ~2.92 Å between structures (2.8× the 1.06 Å noise floor).

Treat V104 / L105 / F107 / A110 / Y355 / L359 as the validated pharmacophore. H187 engagement should not be relied on until these are re-docked into 9XQC.

5. Contents

6. Status and limitations

These are computational designs, not validated compounds. Nothing has been synthesised or assayed.

Citation

Technetium Therapeutics (2026). GPR75 Inhibitor Designs — Technetium GA-II.
Target proposal and binding-pocket modelling: Apodex AI.
Generative design: Technetium TC-43.ai engine.