distillation №23
Tirzepatide — Cys-24 → α-methyl-cysteine (αMe-Cys) substitution; non-canonical Cα-methylated residue replacing the native Cys in the central helix
3D structure
// booting Mol* viewer
// powered by Mol* — drag to rotate · scroll to zoom · use the right panel for cartoon / spacefill / surface presets, measurements & export
AI analysis
tldr
FOLD №23 explores replacing Cys-24 in tirzepatide with α-methyl-cysteine (αMe-Cys), a non-canonical Cα-methylated residue predicted to rigidify the central amphipathic α-helix while eliminating the reactive free thiol. The isolated peptide folds with moderate confidence (pLDDT 0.71), consistent with maintained helical character near the substitution site, but the docked complex against GLP-1R returns a low interface score (ipTM 0.16), leaving receptor engagement geometry unresolved. The heuristic profile reveals a hydrophobic hotspot spanning residues 22–33, consistent with the expected helix region, and the stability score is modest (0.445). The verdict is PROMISING — the backbone conformation rationale is chemically sound, but better docking tools or ensemble prediction would be needed to assess whether αMe-Cys is tolerated at the receptor interface.
detailed analysis
Tirzepatide is a 39-residue dual GIP/GLP-1 receptor agonist that has demonstrated best-in-class glycemic control and weight reduction in the SURPASS and SURMOUNT clinical programs. Its pharmacophore depends critically on an amphipathic α-helix spanning roughly residues 16–30, which docks into the transmembrane bundle of GLP-1R in a manner analogous to native GLP-1. The peptide carries a C18 fatty diacid tether at Lys-20 that drives albumin binding and extends the half-life to approximately five days, but the central helix itself contains a free cysteine at position 24 — an unusual feature for a therapeutic peptide, as free thiols are susceptible to oxidation, disulfide scrambling, and inter-chain crosslinking under oxidative formulation or physiological conditions.
The modification tested in this DISTILLATION replaces Cys-24 with α-methyl-cysteine (αMe-Cys), a non-canonical amino acid carrying an additional methyl group at the α-carbon. This Cα-methylation strategy is grounded in the Thorpe–Ingold steric effect: the gem-disubstitution at Cα significantly narrows the accessible φ,ψ Ramachandran space and biases the residue toward helical conformations. Aib (α-aminoisobutyric acid) is the prototypical example, studied extensively in GLP-1R analog design and explored in FOLD №3 (Retatrutide Aib-2 substitution). αMe-Cys is a close analog that preserves the thioether side chain — critically, it does not change the steric volume or electrostatic character of the side chain at the receptor interface, isolating the conformational backbone effect from any side-chain perturbation. This distinguishes it sharply from FOLD №15, where Glu-16 → homoglutamate in semaglutide extended the side chain and was discarded (pLDDT 0.71), and from the Lys→Arg substitutions in FOLD №10 and prior tirzepatide folds, which modulated side-chain identity and charge.
The structural prediction returned a peptide pLDDT of 0.71 — equivalent to the discarded semaglutide homoglutamate fold (FOLD №15) and the Retatrutide Aib-2 fold (FOLD №3), but below the 0.78 seen in FOLD №10's Lys-17→Arg Retatrutide variant. While 0.71 is not a high-confidence score, it is within an interpretable range for a modified peptide without a canonical Cα at one position, and the structural reasoning agent has flagged this as consistent with maintained central helical character. The pTM of 0.61 is reasonable for the isolated peptide. The critical limitation is the ipTM of 0.16 for the GLP-1R complex, which falls well below the threshold for reliable interface geometry — meaning the docked pose cannot be trusted to report whether αMe-Cys is sterically tolerated in the transmembrane binding cleft or whether it disrupts the Cys-24 packing environment.
The heuristic peptide profile adds texture to the structural result. An aggregation propensity of 0.179 is low, which is pharmacologically favorable — helix-stabilizing modifications that lock amphipathic structure can sometimes expose hydrophobic faces and increase aggregation risk, and this flag is not raised here. The hydrophobic hotspot at residues 22–33 is precisely the region containing αMe-Cys-24 and the downstream Trp-25/Leu-26/Leu-27 cluster, consistent with the amphipathic helix architecture. The stability score of 0.445 is moderate, and the long half-life estimate is expected given the fatty acid tether at Lys-20. The BBB penetration score of 0.051 is irrelevant for a metabolic injectable. No affinity prediction (ΔΔG or binding probability score) was returned from the Boltz-2 affinity module, which is the most significant missing data point for this fold.
The literature context is sparse on the specific question. The retrieved abstracts are exclusively clinical — SURPASS and SURMOUNT trial outcomes — and provide no structural or analog chemistry data on Cys-24's role, tirzepatide's oxidative degradation pathways, or precedent for αMe-Cys in incretin-class peptides. The structural rationale must therefore rest on the broader non-canonical amino acid and GLP-1R structural biology literature, where the general principle (Cα-methylation → helical bias) is well established but the specific tolerance of GLP-1R's transmembrane bundle for a quaternary α-carbon at this position is not characterized. The dual GIP receptor engagement also remains unaddressed — GIPR's tolerance for αMe-Cys at position 24 is an open question.
In aggregate, FOLD №23 earns a PROMISING verdict rather than REFINED because the hypothesis is chemically compelling and the isolated peptide structural signal is interpretable, but the poor ipTM and absent affinity data mean the receptor engagement question — the pharmacologically decisive one — cannot be answered from this run. This is a fold that would benefit substantially from ensemble prediction, a higher-confidence docking tool against the GLP-1R cryo-EM structure, and an explicit ΔΔG estimate. The pharmaceutical rationale for eliminating the free thiol is independently strong and does not depend on binding improvement — even a binding-neutral modification that improves oxidative stability would have formulation value for a molecule dosed weekly at 15 mg.
The broader lab narrative positions this fold as a deliberate evolution in the METABOLIC peptide series. Where earlier tirzepatide folds explored terminal amidation, Lys→Arg side-chain swaps, and N-methylation, FOLD №23 introduces non-canonical amino acid backbone modification — the same category as FOLD №3's Aib-2 Retatrutide experiment, which was discarded for weak signal, but applied here to a mid-helix position with a stronger mechanistic rationale. The Aib substitution at position 2 in FOLD №3 was targeted at DPP-4 resistance at the N-terminus; αMe-Cys-24 targets amphipathic helix rigidity in the receptor-docking domain, a distinct hypothesis with different success criteria.
research data
known activity
// not yet provided by clinical agent
biohacker use
// not yet provided by clinical agent
mechanism class
// not yet provided by clinical agent
AI research brief
Tirzepatide Cys-24 → αMe-Cys: peptide folds with moderate confidence (pLDDT 0.71) and low aggregation risk, supporting helix-locking rationale. Interface score (ipTM 0.16) is too weak to confirm GLP-1R tolerance. Verdict: PROMISING — strong hypothesis, thin structural evidence.
FOLD №23 — Tirzepatide Cys-24 → α-Methyl-Cys: Helix Locking and Thiol Elimination
Verdict: PROMISING | Class: METABOLIC | Target: GLP-1R (P43220)
Mechanism of Action
Tirzepatide is a 39-residue synthetic dual agonist of the glucagon-like peptide-1 receptor (GLP-1R, UniProt P43220) and the glucose-dependent insulinotropic polypeptide receptor (GIPR). Its central amphipathic α-helix (approximately residues 16–30) inserts into the transmembrane bundle of GLP-1R in a conformation analogous to native GLP-1 and semaglutide, stabilizing the active receptor state and driving downstream cAMP signaling, insulin secretion, glucagon suppression, and satiety signaling. The fatty diacid chain at Lys-20 provides albumin binding for extended half-life (~5 days). Position 24 falls within this helix, occupied natively by a free cysteine whose thiol is not engaged in a disulfide bond — an unusual and potentially reactive feature for a therapeutic peptide administered at weekly doses up to 15 mg.
Performance Applications
Tirzepatide is approved for type 2 diabetes (Mounjaro) and obesity (Zepbound), with Phase 3 evidence supporting use in MASH and obstructive sleep apnea. The clinical case for structural optimization of tirzepatide centers on formulation stability and manufacturing robustness rather than potency improvement: the parent molecule already outperforms semaglutide on HbA1c reduction and weight loss in head-to-head SURPASS trials. A more oxidatively stable analog could reduce degradation during long-term storage, simplify formulation without antioxidant excipients, and potentially support higher-concentration presentations for autoinjector delivery. If helical rigidification at position 24 also tightens receptor engagement, a modest potency gain could allow dose reduction with equivalent efficacy — a relevant consideration given dose-dependent GI tolerability profiles.
Modification Rationale
α-Methyl-cysteine (αMe-Cys) introduces a methyl group at the α-carbon of cysteine, creating a quaternary Cα. This Thorpe–Ingold steric effect dramatically restricts the accessible φ,ψ Ramachandran space and enforces helical backbone geometry — the same principle underlying Aib (α-aminoisobutyric acid), the most studied helix-nucleating non-canonical residue in peptide drug design. Critically, αMe-Cys preserves the thioether side chain of cysteine, meaning no change in side-chain volume, charge, or electrostatic character at the receptor interface. This isolates the backbone conformational effect from any side-chain perturbation — a design principle distinct from FOLD №15 (Glu-16 → homoglutamate in semaglutide, which extended the side chain and was discarded at pLDDT 0.71) and from FOLD №3 (Aib-2 in Retatrutide, which replaced the side chain entirely with a methyl group at the N-terminal DPP-4 resistance site).
The elimination of the free thiol is a parallel and independent pharmacological benefit: free cysteines in therapeutic peptides are susceptible to oxidation (sulfenic/sulfinic/sulfonic acid formation), disulfide scrambling with plasma proteins, and crosslinking under oxidative stress. Replacing Cys-24 with αMe-Cys eliminates this liability without introducing a charge, changing the molecular weight significantly, or altering the side-chain contact geometry that may interface with GLP-1R transmembrane residues.
This fold is also conceptually connected to the lab's ongoing exploration of non-canonical amino acids in the METABOLIC class. FOLD №3's Aib substitution in Retatrutide explored helix stabilization at the N-terminus for DPP-4 resistance — a different position and mechanism, but the same chemical category. FOLD №23 applies the Cα-methylation strategy to a mid-helix receptor-docking position for the first time in this lab series.
Predicted Properties (Where Signal Is Moderate)
| Parameter | Value | Interpretation |
|---|---|---|
| pLDDT (isolated peptide) | 0.71 | Moderate confidence — interpretable, not high-confidence |
| pTM | 0.61 | Acceptable fold topology score |
| ipTM (GLP-1R complex) | 0.16 | Poor — receptor interface geometry unreliable |
| Affinity prediction (ΔΔG) | Not returned | Cannot quantify binding change |
| Aggregation propensity | 0.179 | Low — favorable for a helix-stabilized amphipathic peptide |
| Stability score | 0.445 | Moderate |
| Half-life estimate | Long (>6 h) | Expected; dominated by fatty acid tether at Lys-20 |
| BBB penetration | 0.051 | Not relevant for injectable metabolic agent |
The peptide-level pLDDT of 0.71 is consistent with maintained helical character in the central helix region — the predictor does not signal gross structural disruption from the αMe-Cys substitution. The hydrophobic hotspot at residues 22–33 maps precisely onto the amphipathic helix region containing the modification and the downstream Trp-25/Leu-26/Leu-27 cluster, consistent with the expected hydrophobic face of the receptor-docking helix. The low aggregation propensity (0.179) is reassuring — helix-locking modifications that rigidify amphipathic structure can expose hydrophobic faces and promote aggregation, and this flag is not raised here.
The critical signal gap is the ipTM of 0.16. This is too low to support any conclusions about whether αMe-Cys-24 is sterically accommodated in the GLP-1R transmembrane cleft or whether the quaternary α-carbon introduces steric clash with receptor residues. At this interface confidence level, the docked pose geometry should not be interpreted as predictive.
What Would Strengthen This Signal
Additional predictions:
- Ensemble docking against the published GLP-1R–tirzepatide cryo-EM structure (if available) or GLP-1R–GLP-1 complex (PDB: 6X18 and related) using RosettaFold2 or AF-Multimer with multiple seeds, to assess whether ipTM convergence toward ≥0.4 is achievable with repeated runs
- Explicit ΔΔG estimation using a structural perturbation tool (FoldX, Rosetta ddg_monomer, or ProteinMPNN energy estimation) applied to the cryo-EM template with αMe-Cys modeled in at position 24 — this bypasses the ipTM problem by using an experimental receptor structure as scaffold
- GIPR complex docking — tirzepatide's dual agonism requires that any modification be tolerated at both receptors; a parallel fold against GIPR (UniProt P25092) is needed
- Comparative fold against native Cys-24 tirzepatide in this lab's pipeline, to establish a pLDDT/ipTM baseline under identical prediction conditions
Experimental validation (wet lab):
- Solid-phase peptide synthesis of αMe-Cys-24 tirzepatide analog (αMe-Cys is commercially available from Sigma/Bachem as Fmoc-protected building block)
- GLP-1R and GIPR radioligand displacement assay (competitive binding vs. native tirzepatide) — the decisive pharmacological experiment
- cAMP accumulation assay in GLP-1R-overexpressing HEK293 cells — functional potency readout
- Forced degradation study: H₂O₂ stress (0.1–1% w/v, 24 h) comparing native tirzepatide vs. αMe-Cys-24 analog — quantifies the oxidative stability gain independently of receptor binding
- CD spectroscopy in aqueous buffer — direct measurement of helical content change vs. native tirzepatide
What a REFINED verdict would require: ipTM ≥ 0.4 on a repeated or ensembled docking run, or a positive ΔΔG prediction from a structure-based tool, combined with maintained aggregation propensity and no new steric clash flags in the receptor pocket.
folding metrics
// no per-residue pLDDT trace — Boltz-2 returned summary metrics only
aggregation propensity (window)
35 windowsconfidence metrics
domain annotations
// not yet annotated by clinical / structural agents
structural caption
The αMe-Cys24 tirzepatide analog folds into a coherent peptide structure with reasonable per-residue confidence (pLDDT ~0.71), consistent with maintained central helical character around the substitution site. However, the docked pose against GLP-1R shows poor interface confidence (ipTM 0.16), meaning the predicted peptide-receptor contact geometry is unreliable. The heuristic profile flags a hydrophobic hotspot spanning residues 22-33, consistent with the expected amphipathic helix region containing the modification. No affinity prediction was returned, so binding impact cannot be quantified from this run.
peptide profile
These are sequence-based heuristic estimates, not wet-lab measurements. Real aggregation propensity requires TANGO/Aggrescan, real BBB permeability requires QSAR models, and real half-life requires PK studies. Treat the numbers as ranked indicators — useful for comparing variants, not for absolute claims.
known binders
// no ChEMBL binders found for this target
agent findings
caveats
- ─in silico prediction only — requires wet lab validation
- ─single-run prediction (not ensembled)
- ─predicted properties may not reflect real-world biological behavior
- ─this is research, not medical advice
- ─ipTM of 0.16 on the GLP-1R complex is below reliable interpretation threshold — receptor interface geometry should not be treated as predictive
- ─αMe-Cys introduces a quaternary α-carbon not natively present in AlphaFold2/Chai-1 training data; the backbone geometry at position 24 may be imperfectly modeled
- ─no affinity prediction (ΔΔG or binding probability) was returned — binding change versus native tirzepatide is entirely unquantified
- ─heuristic stability, aggregation propensity, and half-life estimates are sequence-based approximations, not experimental measurements
- ─GIPR receptor engagement (required for tirzepatide's dual agonism) was not assessed in this fold — GIP receptor tolerance for αMe-Cys-24 is unknown
- ─the oxidative stability benefit (thiol elimination) is pharmacologically independent of receptor binding but is not directly predicted by structural tools used here
data
works cited
- [1]
(2021). Tirzepatide versus Semaglutide Once Weekly in Patients with Type 2 Diabetes
- [2]
(2021). Efficacy and safety of a novel dual GIP and GLP-1 receptor agonist tirzepatide in patients with type 2 diabetes (SURPASS-1)
- [3]
(2024). Tirzepatide for Metabolic Dysfunction-Associated Steatohepatitis with Liver Fibrosis
- [4]
(2024). The impact of tirzepatide and glucagon-like peptide 1 receptor agonists on oral hormonal contraception
- [5]
(2025). Effect of glucagon-like peptide-1 receptor agonists and co-agonists on body composition: Systematic review and network meta-analysis