Appendix J — A Timeline of Peptide Science

Chapter 3 argued that the history of peptide medicine is not a story about discovering molecules. It is a story about learning to make them, learning to make them survive, and — repeatedly — learning that the rules governing who may sell them matter as much as the chemistry does. This appendix is the evidence for that argument laid out in order.

It is organized into five eras, each with a table of dated events and then prose explaining what changed. The prose is the point. A bare chronology of peptide science reads like a list of achievements and teaches almost nothing; the interesting content is in the gaps between entries, in which problems took decades and which took months, and in how often the binding constant was already known while the drug was still twenty years away.

One honest caveat

A timeline is a selection, and this one is not neutral.

It foregrounds successes, because successes have dates attached and failures usually do not. There is no entry here for the peptide programs that ran for a decade and were quietly shut down, for the compounds that looked excellent in rodents and did nothing in humans, or for the delivery technologies that consumed enormous investment and never reached a label. Those events happened, they outnumber the entries below by a wide margin, and they left no dated record because nobody publishes a press release announcing that a program has been discontinued and the data will not be written up.

Chapter 6's selection argument applies to histories as much as to testimonials. Chapter 6 made the case that a wall of enthusiastic accounts is uninformative if the unenthusiastic accounts were never collected — that you cannot evaluate a numerator without a denominator. A timeline is exactly that shape. It is a numerator. Read the entries below as real events, and hold in mind that the denominator is very large and mostly invisible.

The same caveat applies within entries. Approval dates are crisp and citable; the years of failed formulations that preceded them are not. Nobel Prizes are dated to the year; the disputes about priority and the collaborators who were not included are not. A clean date is often the visible tip of a messy decade, and the cleanliness of the date is a property of the record-keeping rather than of the science.

J.1 Era 1 — Discovering that chemical messengers exist (1900s–1920s)

Before secretin, the dominant model of physiological coordination was nervous. Organs communicated through nerves; that was the mechanism, and it was a good mechanism that explained a great deal. When Bayliss and Starling demonstrated that a substance released from the intestinal lining could travel through the bloodstream and cause the pancreas to secrete — with the relevant nerves cut, which was the entire force of the experiment — they established something genuinely new. The body has a second signaling system, chemical rather than electrical, broadcast rather than wired.

That is the conceptual foundation everything else in this book stands on. A peptide drug works because the body already runs on chemical messengers and a peptide drug is one, or resembles one closely enough to be read as one by a receptor. Without the secretin result there is no framework in which "give the patient a signaling molecule" is even a coherent idea. Naming the category three years later was not a footnote either. Starling's word gave a scattered set of observations — about the pancreas, the thyroid, the adrenals — a single heading, and fields with headings attract work in a way that collections of curiosities do not.

Insulin is the event that made the field consequential, and it deserves to be understood as unusually fast rather than as typical. From isolation to the first treated patient took months. The Nobel Prize followed within about a year of that. Nothing else in this appendix moves at that speed, and Chapter 3 was emphatic about why: the target disease was rapidly fatal in young patients, the effect of the treatment was unmistakable within hours, and there was no need for a statistical argument because the clinical change was visible to anyone in the room. Insulin did not need a trial design to be convincing. Almost everything after it did.

It is worth registering what was not known in 1922. Nobody knew insulin's sequence. Nobody knew it was a peptide in any structural sense that would satisfy a modern chemist. Nobody could make it except by extracting it from animal pancreas, which meant supply was a slaughterhouse logistics problem and purity was whatever the extraction happened to deliver. The drug preceded the chemistry by thirty years, which is the reverse of how the modern pipeline is imagined to work and a useful corrective to the assumption that understanding a molecule is a prerequisite to using it.

J.2 Era 2 — Learning what peptides are made of (1950s–1960s)

Sanger's result is easy to underrate because its conclusion now sounds obvious. Before it, there was serious doubt that proteins had defined sequences at all. A reasonable competing view held that proteins were statistical assemblies — mixtures with characteristic compositions and variable orders, more like a polymer blend than like a text. Sanger's insulin sequence settled it: this protein has one sequence, the same one every time, and it can be written down.

That single fact is the premise of Chapter 1 and therefore of the entire book. If proteins were statistical, "the sequence of GLP-1" would be a category error, Appendix B would have nothing to tabulate, and the engineering program of Chapter 33 — change residue 8, change residue 34, attach at residue 26 — would be unthinkable. Defined sequence is what makes rational modification possible. It also, incidentally, is what makes analytical identity testing possible: Chapter 34's arguments about what a certificate of analysis does and does not establish presuppose that there is a single correct answer for the instrument to find.

Du Vigneaud's oxytocin did something Sanger's insulin could not. Sanger determined a structure by taking a molecule apart; du Vigneaud confirmed one by putting a molecule together and showing that the synthetic material had the same biological activity as the natural hormone. Synthesis as proof is a stronger form of evidence than analysis, because it forecloses the possibility that the proposed structure is merely consistent with the data. Oxytocin is nine residues with one disulfide (Appendix B §B.4) — small enough to be tractable with the methods of the time and biologically active enough to make the demonstration convincing.

It also opened a door that this book walks through repeatedly. If a hormone can be built rather than extracted, then it can be built differently — with a substitution, with a D-residue, with a modified terminus. Every analog in Chapter 21 and Chapter 27 descends conceptually from the moment a peptide hormone became something a chemist could make.

Merrifield's contribution is the one with the largest practical footprint, and Chapter 32 spent a chapter on it for good reason. Before solid-phase synthesis, building a peptide meant coupling residues in solution and purifying the growing chain after every step — a process where yields compounded downward and the labor per residue was punishing. Anchoring the chain to an insoluble resin changed the economics completely: excess reagents can be washed away rather than separated from, coupling can be driven to completion by simple excess, and the whole cycle becomes repetitive enough to automate.

Merrifield did not discover a molecule. He removed a manufacturing bottleneck, and the consequences were enormous — automated synthesizers, peptides available to any lab that wanted them, and eventually the industrial capacity that makes a peptide drug at population scale possible. It is the first entry in this appendix that fits the pattern the closing section identifies: the advance that changed the field was about making rather than about knowing. Note also the gap between the introduction and the prize, which is roughly two decades. That gap is not unusual and it is worth carrying as a calibration point for how long it takes a method's importance to become undeniable.

J.3 Era 3 — Mapping the signaling systems (1970s–1980s)

This is the era in which the map got drawn. Guillemin and Schally's work — extracting vanishingly small quantities of hypothalamic peptides from enormous quantities of tissue — established that the brain's control over the endocrine system is exercised through peptide messengers. TRH, GnRH, and somatostatin are all in this book: GnRH underlies the -relin and -relix compounds of Appendix I §I.5 and Chapter 27; somatostatin is the parent of octreotide and, eventually, of the radionuclide therapy in Era 4. The hypothalamic peptides are the reason "peptide" and "hormone axis" are nearly synonymous in clinical usage.

The shared half of that prize deserves equal billing. Radioimmunoassay made peptides measurable at the concentrations they actually circulate at, which are far below what earlier chemistry could detect. Nearly every quantitative statement in Chapters 7 through 14 — about circulating levels, about half-lives measured in minutes, about what happens to a hormone after a meal — rests on assay technology that did not exist before this. Appendix B §B.4 notes that radioiodination at tyrosine was the standard labeling route; that is this technology's fingerprint on the field. A discovery era and a measurement era are the same era, and that is not a coincidence.

The enkephalin identification opened the endogenous opioid system that Chapter 20 examines, and it carries a lesson the book returns to. The receptors were characterized before the natural ligands were found — pharmacology had established that the brain contained binding sites for plant-derived opioids, which implied that the brain made something of its own to fill them. The peptides were found by looking for what the receptor was waiting for. This is the discovery logic that recurs throughout peptide science: find the receptor, deduce that a ligand exists, go find it.

Three entries in this era are really about supply, and together they make Chapter 3's central point. Recombinant DNA production of human insulin ended the dependence on animal pancreas — a supply chain that put a hard ceiling on how many patients could be treated and made the drug's availability a function of meat consumption. The approval of recombinant human insulin as the first approved recombinant DNA drug established the regulatory pathway that essentially every protein therapeutic since has traveled.

And then the growth hormone entry, which is the darkest item in this appendix and the one Chapter 3 and Chapter 14 both insist on. Human growth hormone was, until this point, extracted from pituitary glands taken from cadavers. That sourcing transmitted the agent that causes Creutzfeldt-Jakob disease to some recipients — a fatal neurodegenerative illness, in people who had been treated as children, appearing after a latency measured in years. The cadaver-derived material was withdrawn, and recombinant human growth hormone became available in the same year.

Read those two facts together, because their simultaneity is the lesson. The technology that made the dangerous product unnecessary arrived at almost exactly the moment the danger became undeniable. It would be comfortable to conclude that the system worked. The more accurate conclusion is that the harm had already been done, that it had been done through a sourcing decision rather than a pharmacological one, and that nobody involved had reason to suspect the mechanism — prion biology was not understood, and the extraction process was not designed to remove something nobody knew was there. Chapter 14 uses this as the definitive argument for why the origin of a peptide preparation is a safety question independent of the molecule's own properties. A correct sequence, correctly synthesized, is a different object from a correct sequence extracted from unscreened human tissue, and no analysis of the molecule distinguishes them.

Two forward-looking entries close the era. Captopril — not a peptide — was developed from bradykinin-potentiating peptides found in pit viper venom, working through the peptide teprotide to a small molecule that could be taken orally. It is the first major peptidomimetic: a demonstration that a peptide can be a lead rather than a destination, and that the useful thing about a peptide is sometimes the shape of its interaction rather than the molecule itself. Chapter 35 treats venom as a discovery resource; captopril is the proof of concept that made that a serious research program rather than a curiosity.

Phage display did something structurally similar for discovery. Instead of finding peptides by extracting them from tissue, it made it possible to search vast libraries of sequences for ones that bind a chosen target. Chapter 35 covers the method; here the point is that it converted peptide discovery from a natural-products problem into a screening problem, which is a different kind of activity with a different rate of progress.

Finally, the identification of GLP-1 as an incretin hormone, emerging from work on how the proglucagon precursor is processed differently in different tissues. Chapter 7 tells this story properly. For the timeline, note only the date, and hold it: this is where the clock starts for the compounds that dominate Era 5.

J.4 Era 4 — Peptides become mainstream drugs (1990s–2010s)

The exendin-4 story is the one people remember, and it deserves its status. A GLP-1 analog that resists DPP-4 — solving, in a single naturally occurring molecule, the survival problem that had made native GLP-1 useless as a drug — was found in the venom of a lizard. Chapter 35 draws the general lesson about venoms as a source of highly optimized bioactive peptides. Chapter 33 draws the narrower one: evolution had already built the modification the medicinal chemists were looking for, and finding it took a decade less than designing it would have.

Ziconotide, from a cone snail, makes the counterpoint that keeps venom-derived optimism honest. It works, it is approved, and it is delivered intrathecally — directly into the cerebrospinal fluid, because it cannot reach its target any other way. Chapter 4's delivery constraints do not relent for interesting molecules. A peptide can be potent, selective, and clinically real and still require a route of administration that limits it to a narrow population. Potency is not practicality, and the gap between them is where most peptide programs die.

The GLP-1 sequence in this era is the one to trace carefully, because Era 5 is unintelligible without it. Exenatide arrives as the first agonist in the class — roughly two decades after the hormone was identified as an incretin. Liraglutide follows five years later, carrying the lipidation strategy that Chapter 33 dissects: a fatty acid attached to the peptide, binding albumin, extending half-life from minutes to about a day. Semaglutide follows liraglutide by seven years with a more aggressive version of the same strategy — a different fatty acid, the Aib substitution at position 8, and the Lys→Arg substitution at position 34 that Appendix B §B.4 explains — reaching a half-life that supports weekly administration.

Notice what changed across those three molecules and what did not. The receptor is the same. The biology being engaged is the same. The pharmacology of GLP-1 receptor agonism was well understood before the first of them was approved. What changed, each time, was how long the molecule survives in the body — and that is what converted an interesting hormone into a drug people actually take. This is the pattern the closing section names, and this era is where it is most visible.

The liraglutide entries also demonstrate something Appendix I §I.6 discussed. The 2010 and 2014 approvals are the same molecule. Different dose, different indication, different trial package, different trade name. Popular coverage regularly discusses these as though they were two drugs, and comparative claims built on that footing are comparing doses rather than chemistry.

Two entries in this era are regulatory rather than scientific, and they are the two with the largest downstream footprint. DSHEA created the U.S. dietary supplement framework — a category with different requirements from drugs, different premarket obligations, and different rules about what may be claimed. Chapter 38 works through why so many peptide-adjacent products are sold into or alongside that category, and why the boundary is contested. The Drug Quality and Security Act established the 503A and 503B compounding framework: the rules under which pharmacies and outsourcing facilities may prepare medications not commercially available in the needed form. Chapter 12 covers compounding directly. Both entries look procedural on the page. Both determine, more directly than any laboratory result, what a person can actually obtain.

The two remaining entries broaden the picture in useful directions. Lutetium-177 dotatate uses a somatostatin analog as a targeting device — the peptide carries a radionuclide to tumors that overexpress the somatostatin receptor, and the radiation does the work. Chapter 27 covers it. It is the clearest case in the book of a peptide functioning as an address rather than as a signal, and it descends directly from Guillemin and Schally's somatostatin four decades earlier. Bremelanotide is worth including precisely because it is not metabolic: a melanocortin receptor agonist approved for a sexual desire indication (Chapter 24), a reminder that the peptide drug class is not coextensive with the GLP-1 story that currently dominates attention.

Oral semaglutide deserves its asterisk. It is a genuine achievement — an orally administered peptide, co-formulated with SNAC, an absorption enhancer that transiently permits uptake across the gastric mucosa. Chapter 4 explains the mechanism and Chapter 36 explains the limits: oral bioavailability by this route is low and variable, which is managed through dose and administration conditions rather than eliminated. It is a real solution to Chapter 4's central problem and it is not a general one. A single successful oral peptide does not mean peptides are now orally available, and the distance between those two statements is where a lot of marketing lives.

J.5 Era 5 — The metabolic revolution and the computational turn (2020s)

This era has two separate stories running in parallel, and they barely touch.

The first is metabolic. Semaglutide's approval at a higher dose for weight management is the moment peptide drugs became a subject of general conversation rather than a specialist topic. Tirzepatide followed with a genuine mechanistic addition — dual agonism, engaging both the GIP and GLP-1 receptors with one molecule, which Chapter 9 examines as the first commercially decisive demonstration that multi-receptor peptides can outperform single-target ones. Both then acquired weight-management indications after their diabetes indications, repeating the two-indication pattern liraglutide had established.

SELECT is the entry to read most carefully, because it is the one most often reported badly. In roughly 17,000 participants with established cardiovascular disease and overweight or obesity, and without diabetes, semaglutide reduced major adverse cardiovascular events by about 20% relative. The absolute change was roughly 8% to 6.5% — about 1.5 percentage points.

Both of those numbers are correct and they describe the same result. Chapter 10 works through why you need both. The relative figure is the right one for asking whether the drug does something to cardiovascular biology; the absolute figure is the right one for asking what it does for a given person. A 20% relative reduction sounds like it belongs in a different category of magnitude from a 1.5-percentage-point absolute reduction, and the gap between how those two sentences feel is the single most reliable source of miscalibration in medical reporting. The result is real, it is substantial for a population at that level of baseline risk, and it is not what "cuts heart attacks by twenty percent" conveys to a general reader. A cardiovascular indication was approved on that basis the following year — an event worth noting because it extended the drug's formal use beyond glycemic control and beyond weight.

The second story is computational, and it is genuinely a step change. AlphaFold2's CASP14 result ended a decades-old open problem in structural biology: predicting a protein's three-dimensional structure from its sequence, reliably, without solving it experimentally. The Nobel Prize four years later recognized both that prediction work and David Baker's computational protein design — the inverse problem, of specifying a desired structure and computing a sequence that folds into it. Chapter 35 covers what this means for peptide discovery.

It also covers what it does not mean, and the timeline is the right place to be blunt about it. Structure prediction and design address the target-and-binding half of drug development, which Era 4 demonstrated was rarely the bottleneck. Knowing a structure does not tell you whether a molecule survives in plasma, whether it can be manufactured at scale, whether it crosses any membrane it needs to cross, whether it is immunogenic, or whether engaging its target produces a clinical benefit in people. Those are the questions that consumed the decades in every earlier era. Computational methods are a real and large acceleration of one stage. Read claims that they will compress the whole timeline against the rest of this appendix.

The regulatory entries in this era are, as usual, the ones with the most immediate effect on what people can obtain. The insulin transition — under which insulin was deemed a biological product in the United States — reclassified a century-old drug, opening the biosimilar pathway to it and changing the competitive landscape for a product that had been regulated as a conventional drug for its entire commercial life. Chapter 38 §38.2 covers the mechanics. It is a clean illustration of a point that recurs: the molecule did not change at all, and what could be sold changed completely.

The BPC-157 entry says something specific and limited, and it is easy to over-read. Placing a compound on the WADA Prohibited List under category S0 — non-approved substances — is a statement that no regulatory authority has approved it for human therapeutic use, and that it is therefore prohibited in sport at all times. It is not a finding that the compound works, and it is not a finding that it is dangerous. Chapter 38 discusses the S0 category directly. Appendix I §I.4 made the adjacent point about name form; this entry is the same observation arriving from a different institution. A compound can be simultaneously banned in sport, sold widely online, and unsupported by any completed human outcome trial, and those three facts are compatible because they are about three different things.

The final entry is deliberately undated, and the reason is a lesson in itself. Widespread shortages of GLP-1 receptor agonists in the mid-2020s created conditions under which compounded versions could be prepared and sold at very large scale under the 503A and 503B framework established in Era 4; subsequent declarations that the shortages had resolved restricted that market. The details and timing were contested — through litigation, through disputed determinations, and through disagreement about what "resolved" meant in practice for patients who still could not fill prescriptions. Chapters 12 and 38 lay out the structure of the dispute.

Giving that entry a false-crisp date would misrepresent it. It was not an event; it was a period, with different start and end dates depending on the product, the region, and who was doing the counting. A timeline that dates everything to the day teaches its reader that everything is datable to the day, and that is a bad habit to acquire. Some things in this field happened on a date. Others happened over a couple of years while people argued about whether they were happening.

J.6 What the timeline shows

Four observations. Each one is an argument this book makes elsewhere; the timeline is what makes them checkable rather than assertable.

One: the gap between discovery and drug is measured in decades

Insulin is the fast case — months from isolation to a treated patient — and it was extraordinary rather than representative. It was fast because the disease killed quickly, the effect was visible immediately, and no statistical apparatus was required to see it. Nothing else in this appendix resembles it.

Look instead at GLP-1. Identified as an incretin hormone in the mid-to-late 1980s. First receptor agonist approved in 2005 — roughly two decades later. The molecule that made the class a subject of general conversation, approved at a weight-management dose in 2021, arrived more than a decade after that. Between the identification and the household name lies something on the order of thirty-five years, populated by delivery failures, half-life problems, manufacturing scale-up, and the slow accumulation of outcome data.

That is the timescale to judge "coming soon" claims against. Chapter 36 asked readers to interrogate pipeline enthusiasm with specific questions: what phase, which sponsor, which registered trial, what has been published. The timeline supplies the prior those questions operate against. A compound described as being "a few years away" is describing an outcome that, historically, has taken a decade or more from a much more advanced starting point than most such compounds occupy — and that is conditional on the program succeeding, which most do not.

Two: almost every major advance solved a delivery or survival problem, not a target problem

Walk the entries and ask what each one fixed.

Sanger and du Vigneaud established what the molecules are. Merrifield made them buildable. Recombinant DNA made them affordable and safe to source. Lipidation — the strategy running through liraglutide and semaglutide — made them last. SNAC made one of them orally absorbable, partially. Cyclization, D-substitution, and the Aib substitution at position 8 all address the same category of problem: the molecule works and then it is destroyed too quickly to matter.

Against that list, set the target-side entries. The GLP-1 receptor was understood well before any agonist was approved. The somatostatin receptor was characterized decades before lutetium-177 dotatate targeted it. The opioid receptors were characterized before the enkephalins were found. In case after case, the receptor pharmacology was known and the drug was still twenty years away, because knowing what a molecule should bind is a different problem from getting a molecule to the binding site intact and keeping it there.

Chapter 33 §33.10 stated this as the engineering thesis of the book, and this appendix is its historical evidence. The receptor pharmacology was rarely the bottleneck. When you evaluate a claim that a new compound will be transformative because of what it binds, the timeline's answer is that binding was almost never the hard part.

Three: the regulatory milestones sit alongside the scientific ones and are not decoration

Four entries in this appendix are legal rather than laboratory events: the 1938 Act that established the modern U.S. drug regulatory framework, DSHEA in 1994, the 2013 compounding framework, and the 2020 insulin transition. A reader skimming for science would skip all four.

Each one changed what could be sold and to whom, and each did so faster than any laboratory result did. DSHEA created a category, and an industry organized itself around that category's boundaries within a few years. The compounding framework set rules that, a decade later, determined the shape of the largest gray market in this book's subject matter. The insulin transition reclassified a century-old molecule and changed its competitive landscape without a single experiment being performed.

Chapter 38 argued that a reader who understands the pharmacology and not the regulation will misunderstand the market they are actually looking at — will assume that availability implies approval, that a product's existence implies oversight, that a compound sold in a vial went through something resembling the process that put semaglutide in a pen. The timeline shows why that assumption fails: the rules governing availability have their own history, moving on its own schedule, and it intersects the science less often than you would expect.

Four: the compounds with the longest histories are the best evidenced, and the correlation is not a coincidence

Insulin has a century behind it — a century of formulations, of outcome studies, of registries, of observed rare adverse effects, of use in populations no trial would have enrolled. The GLP-1 receptor agonists have two decades, including large cardiovascular outcome trials that took years to run and enrolled tens of thousands of people. These are the best-evidenced peptides in the book, and they are also the oldest.

Meanwhile, several compounds discussed online with total confidence have no completed human outcome trial at all. Appendix A's ratings track this closely enough that the correlation is hard to miss.

Be careful about the direction of the inference. Time in the world is not itself evidence. A compound that has circulated for twenty years without being studied has accumulated exposure, not data, and Chapter 6's argument about uncontrolled experience applies in full: a great many people using something and reporting on it does not generate the comparison that would tell you whether it works.

But the accumulation of evidence takes time, necessarily and irreducibly. An outcome trial takes years to enroll and years to follow. Rare adverse effects require large denominators and long observation. Confidence in a therapy is a function of accumulated study, and accumulated study is a function of elapsed time multiplied by sustained investigation. The second factor is the one that varies.

Which yields the practical reading, and it is Chapter 36 §36.10's: a compound that has been "promising" for fifteen years is telling you something. Not that it is useless — occasionally a good compound stalls for reasons of funding, patent economics, or bad luck. But the base rate is unkind. Fifteen years of promise with no completed trial usually means the trials were not run, and the most common reason trials are not run is that nobody with the resources to run one believed the result would justify the expense. The absence of evidence, sustained across a period in which evidence could easily have been generated, is weak evidence of absence — and it is very often the most informative fact available about a compound whose enthusiasts insist the science is settled.

Related: Chapter 3 (history) · Chapter 4 (delivery) · Chapter 6 (evidence and selection) · Chapter 7 (incretins) · Chapter 8 (GLP-1 agonists) · Chapter 9 (dual agonists) · Chapter 10 (cardiovascular outcomes) · Chapter 12 (compounding) · Chapter 14 (growth hormone) · Chapter 20 (opioid peptides) · Chapter 27 (oncology peptides) · Chapter 32 (synthesis) · Chapter 33 (engineering) · Chapter 35 (discovery) · Chapter 36 (pipelines) · Chapter 38 (regulation) · Appendix A (evidence ratings) · Appendix B (amino acid reference) · Appendix I (nomenclature) · Appendix K (glossary)