Chapter 33 — Key Takeaways
The one sentence
Peptide engineering solves delivery problems and only delivery problems — it gets the molecule to the receptor intact, in enough quantity, for long enough, by an acceptable route, in a form that can be manufactured reproducibly. What happens after the molecule arrives is biology, and biology is not a chemistry problem.
The framework
Chapter 4's two exits organize everything. A peptide leaves the body by proteolysis and by renal filtration. Both must be closed. "Protease-resistant" is half a claim, not a whole one.
Four goals, not a list of tricks. Every modification in the chapter serves one or more of: resist degradation, increase effective size, improve selectivity or potency, or enable a route.
Goals conflict. The features an enzyme recognizes and the features a receptor recognizes overlap far more often than anyone would like. Most engineering failures are a modification that worked beautifully for stability and cost most of the potency.
Nothing in the toolkit makes a peptide better at its job. For the great majority of peptide drugs the natural ligand was already excellent at the receptor. Intrinsic pharmacology was rarely the bottleneck.
The techniques, compressed
| Technique | What it defeats | Signature example |
|---|---|---|
| Aib substitution | a specific protease cut site | semaglutide position 8 |
| D-amino acids | proteases generally | octreotide, cyclosporine, GnRH antagonists |
| Cyclization | exopeptidases; also improves affinity | octreotide, cyclosporine |
| Stapling | protease attack; helix instability; contested cell entry | p53–MDM2 programs |
| Lipidation | renal filtration (and proteolysis) | semaglutide, liraglutide, insulin detemir |
| PEGylation | renal filtration | PEGylated interferons, uricase — now displaced for peptides |
| Fc / albumin fusion | filtration, via size + FcRn recycling | dulaglutide, romiplostim |
| Peptidomimetics | the entire peptide liability set | orforglipron |
Aib forces chemical synthesis. No codon, no tRNA, no ribosome. One non-proteinogenic residue determines the whole manufacturing route (Chapter 32).
Cyclization improves affinity for an entropic reason — it pre-pays the conformational cost of binding — not by adding contacts.
Lipidation is tuned, not merely applied. Chain length, a diacid terminus, and spacer chemistry are three independent knobs. The same tool produced liraglutide (~13 hours), semaglutide (~1 week), and insulin detemir (a flat basal profile).
Fusion means leaving the category. Very long half-life, biologic manufacturing, biologic costs, biosimilar rather than generic competition.
Peptidomimetics are the endpoint of the arc — the ultimate modification to a peptide is to stop using one.
The anchor: semaglutide in three lines
| Change | Position | Problem solved |
|---|---|---|
| Ala → Aib | 8 | proteolysis — abolishes the DPP-4 cut site |
| Lys → Arg | 34 | manufacturing — one reactive amine, so acylation gives one defined product rather than attachment isomers. No patient benefit. |
| C18 diacid via spacer | 26 | renal filtration — reversible albumin binding |
Half-life ~1–2 minutes → ~1 week. Receptor activity unchanged.
The receptor pharmacology was never the problem; evolution had optimized it. What was engineered was survival.
Multi-agonist architectures
One molecule, two receptors (tirzepatide): ratio fixed by structure, one pharmacokinetic profile, one stability and immunogenicity program. Changing the ratio requires a new molecule.
Two molecules, coformulated (CagriSema): ratio adjustable by formulation, two pharmacokinetic profiles that can diverge, two stability programs, two immunogenicity questions.
Neither is obviously better. A head-to-head clinical result is evidence about two products, not a verdict on two architectures.
What engineering cannot fix
Chapter 2 §2.7 named four termination mechanisms:
- Degradation — property of the molecule → engineerable
- Clearance — property of the molecule → engineerable
- Desensitization — decision of the target cell → not engineerable
- Downregulation — decision of the target cell → not engineerable
And worse: a longer-lasting agonist pushes harder on exactly the machinery that shuts the signal down. The native pulse is short partly so that adaptation never fully engages.
Engineering also cannot fix a wrong target (Chapter 22, substance P), an unfavorable therapeutic window, a surrogate endpoint that does not predict outcome (Chapter 16), or the absence of clinical evidence.
Everything in Part III that fails, fails somewhere else.
Ratings issued in this chapter
| Claim | Rating |
|---|---|
| Lipidation / albumin binding as a half-life extension strategy | ✅ |
| Stapled peptides achieve meaningful intracellular delivery | ⚠️ |
| Claim form — "this modification makes the peptide more effective" | ❌ (as usually stated) |
| Peptidomimetics / small-molecule agonists at peptide GPCRs as a viable route | ⚠️ |
Dossier
Field 4 is a modification map: native reference, native half-life, changes from native, the problem each change solves, resulting half-life and its source, and what remains unsolved.
For many gray-market compounds the honest entry is "unmodified native sequence or fragment." That entry is not an accusation — it is a prompt for two questions answerable without any clinical evidence: what is its half-life, actually? and does the way it is used make sense against that number?
The three questions to carry forward
- Which exit does this close — and what about the other one?
- Did this change duration and delivery, or did it change what the molecule does? (Almost always the former.)
- If this failed, did it fail at delivery — or somewhere engineering cannot reach?