From pig pancreas to AlphaFold: 100 years of peptide therapy in five inflection points

Peptide therapy is not a single technology — it is a hundred-year arc of breakthroughs that compound on each other. Each step removed a different bottleneck. Five moments matter most.

1922 — Insulin extracted

Banting, Best, Macleod, and Collip extracted insulin from animal pancreas tissue at the University of Toronto. Within a year, type 1 diabetes — a death sentence in 1921 — became a manageable chronic condition. Insulin was the first peptide drug. It worked because animal insulin is structurally close enough to human insulin that the human receptor accepts it, with the famous trade-off of antibody response over time.

1963 — Solid Phase Peptide Synthesis

R. Bruce Merrifield invented Solid Phase Peptide Synthesis at Rockefeller University. Before SPPS, every amino acid you added to a chain required a separate purification, and a complex sequence took years. Merrifield anchored the growing peptide to an insoluble resin so excess reagents could simply be rinsed away after each step.

Reaction efficiency jumped to ~99.5% per coupling step, and timelines collapsed from years to days. Every modern peptide drug — semaglutide, tirzepatide, BPC-157, every one — is downstream of this single breakthrough. Merrifield received the 1984 Nobel Prize for it.

Merrifield’s insight was simple: anchor the chain so the bench can rinse instead of purify. It is the move that turned peptide chemistry into peptide manufacturing.

1982 — Recombinant human insulin

Genentech and Eli Lilly produced the first recombinant human insulin (Humulin) using genetically engineered E. coli. The animal-extract era ended. Industrial-scale production of a drug whose protein sequence exactly matched the human version became possible, removing the antibody-response problem that had plagued animal insulin for sixty years. Recombinant DNA had its first blockbuster.

2017 — Semaglutide approved

Novo Nordisk’s once-weekly semaglutide (Ozempic) was approved for type 2 diabetes. In trials it produced ~15% body-weight loss as a side effect, which is what eventually made Wegovy (the obesity-indicated dose) the most disruptive peptide drug since insulin. The molecule did not invent a receptor; it engineered an existing gut hormone (GLP-1) for stability and albumin binding using three modifications on a 31-residue scaffold.

Tirzepatide followed in 2022 with a dual GLP-1/GIP agonist mechanism and ~20–25% weight loss in trials. The "GLP-1 economy" that defines the field today is the direct downstream of these two approvals.

2024+ — AI as a design partner

AlphaFold3 (DeepMind) and RFdiffusion (Baker Lab) made high-confidence 3D-structure prediction and generative backbone design routine. Toxicity screens like ToxGIN and tAMPer catch hemotoxicity in silico before any wet-lab work. Protein language models — ESM-2, PepMLM — apply transformer architectures to biology, generating context-aware sequences. Agentic-reasoning models like PepThink-R1 add chain-of-thought scientific reasoning so design failures are traceable.

The frontier challenge is the metrics-reality gap: a peptide can look ideal in silico and still fail in a real human body because of metabolic instability, immunogenicity, or microenvironment effects. Closing that gap is the next inflection point — the one that has not happened yet.