FOLD №15 — Semaglutide Glu-16 → Homoglutamate

Verdict: DISCARDED | Target: GLP-1R (P43220) | Class: Metabolic

Mechanism of Action (Background)

Semaglutide is a GLP-1 receptor agonist that activates GLP-1R, a class B GPCR expressed on pancreatic β-cells, gut enteroendocrine cells, and the hypothalamus. Agonism drives glucose-dependent insulin secretion, glucagon suppression, gastric emptying delay, and centrally mediated appetite reduction. Its therapeutic value in T2DM and obesity management (STEP and SUSTAIN trial programs) is well established. The peptide's pharmacological identity rests on two structural pillars: Aib-8 (DPP-4 resistance) and a C18 fatty-diacid at Lys-26 (albumin binding, ~46 h half-life in mini-pigs). These features were deliberately optimized at a modest cost to raw GLP-1R affinity — semaglutide binds at ~0.38 nM, approximately 3-fold weaker than liraglutide — suggesting that affinity headroom exists in the scaffold.

Modification Hypothesis (What We Tested)

Fold №15 replaced Glu-16 in the central α-helix of semaglutide with homoglutamate (Hge), a β-methylene-homologated glutamate that extends the carboxylate reach by approximately 1.5 Å. The hypothesis: cryo-EM structures of GLP-1 analogs bound to GLP-1R show Glu-16 making an acidic contact with ECD Arg310/Arg380; if the native glutamate sits marginally outside optimal salt-bridge geometry (~4 Å), homoglutamate's extended reach (~3 Å) could tighten this electrostatic contact and improve receptor affinity. The modification is charge-neutral (preserves carboxylate and pKa), is expected to be helix-compatible (β-homologated amino acids are well tolerated in α-helical contexts), and is chemically distinct from both the DPP-4 resistance element at position 8 and the albumin tether at position 26.

This strategy differs meaningfully from prior lab folds: Fold №3 (Retatrutide Aib-2) tested backbone rigidification for DPP-4 resistance; Fold №10 (Retatrutide Lys-17→Arg) tested intra-helix i,i+4 salt-bridge optimization. Fold №15 is uniquely focused on inter-molecular reach extension at a receptor-facing interfacial residue — a distinct design axis.

Why the Prediction Was Uninformative (Technical Analysis)

The prediction pipeline returned the following outputs:

The pLDDT of 0.714 and pTM of 0.783 are respectable scores, consistent with semaglutide's known strong α-helical propensity. However, these scores characterize the peptide backbone in isolation — they do not reflect a receptor-docked complex and cannot address salt-bridge geometry at the GLP-1R ECD interface. The central failure of this fold is the absence of Chai-1 agreement and Boltz-2 affinity values: without cross-model consensus on a co-folded complex and without an affinity delta, the hypothesis that Hge extends productive electrostatic reach cannot be evaluated.

This is a tool limitation, not a chemistry failure. The semaglutide scaffold is a large, lipidated, modified peptide (31 residues + non-canonical Aib + fatty-acid tether); the prediction of its full receptor-bound complex is at the edge of current single-run pipeline capability. Fold №3 (Retatrutide Aib-2) encountered a comparable outcome at pLDDT 0.71 — suggesting this class of highly modified metabolic peptides consistently challenges single-run predictors at the interface level. By contrast, Fold №10 (Retatrutide Lys-17→Arg, pLDDT 0.78) produced a PROMISING verdict, likely because the intra-helix salt-bridge geometry is more accessible to structure prediction than an inter-molecular peptide–receptor contact.

Heuristic sequence-based estimates provided:

• Aggregation propensity: 0.152 (low — expected for a designed helix-forming analog)
• Stability score: 0.432 (moderate — consistent with modified peptide class)
• BBB penetration: 0.024 (negligible — appropriate for a large peripheral agonist)
• Half-life: moderate-to-long (reflects albumin-tethered class behavior)

None of these heuristics bear on receptor affinity or salt-bridge geometry.

What This Tells Us (Negative Results Are Data)

This fold's discarded status is informative in three respects:

1. The pipeline cannot currently adjudicate inter-molecular side-chain reach hypotheses for lipidated class B GPCR peptides in a single run. The absence of Boltz-2 affinity output and Chai-1 consensus is a consistent challenge for this peptide class. Future folds testing interfacial contacts on semaglutide or similar analogs should plan for ensemble runs or co-crystal-seeded docking from the outset.

2. The chemistry is not ruled out. The Hge substitution has not been predicted to be destabilizing, aggregation-prone, or conformationally disruptive. pLDDT 0.714 and pTM 0.783 are not failure scores — they indicate a well-folded backbone. The biological hypothesis remains structurally plausible; it is simply unresolved by these tools at this time.

3. The semaglutide scaffold's affinity headroom (0.38 nM vs. liraglutide's ~0.12 nM) remains an open optimization target. This fold does not close that question. It only establishes that a single-run AlphaFold pipeline without usable complex geometry is insufficient to evaluate it.

Alternative Hypotheses to Test (Avoid the Failure Mode)

1. Ensemble docking with GLP-1R ECD co-crystal input: Use the published GLP-1 peptide–GLP-1R ECD structure (PDB: 5NX2 or equivalent) as a structural prior for Rosetta or Glide docking of the Hge-16 variant. This bypasses the cold-start complex prediction problem entirely.

2. Molecular dynamics on the Glu-16/Arg310 contact: Short (50–100 ns) MD simulation of native semaglutide vs. Hge-16 variant in the receptor-bound pose would directly quantify salt-bridge occupancy and distance distributions — the exact metric the hypothesis requires.

3. Paired Glu-16 alanine scan first: Before homologating, computationally testing Glu-16 → Ala (a charge-ablation control) would confirm whether the position contributes to binding affinity at all in the in silico framework. If Ala substitution shows predicted affinity loss, the salt-bridge hypothesis is supported; if neutral, the position may not be a productive optimization target.

4. Position 12 or 13 acidic residue extension: If Glu-16's receptor contact is difficult to validate, equivalent homologation at other acidic positions in the helical face (Asp15 or Glu22 in related GLP-1 scaffolds) could be tested where cryo-EM data more clearly resolves the ECD contact geometry.

5. Synthesize and test directly: Given the chemical conservatism of the Hge substitution and the low aggregation propensity predicted (0.152), this analog is a reasonable synthesis target for a cAMP accumulation assay against GLP-1R — the wet-lab answer is not gated on a successful in silico run.

⚠️ Mandatory disclaimer: All results are in silico predictions only. No wet-lab validation has been performed. Predicted properties do not reflect confirmed biological activity. This is computational research, not medical advice. Heuristic property estimates are sequence-based approximations, not experimentally measured values.