The Price of a Gram Over Sixty Years

In a laboratory at Cornell in the early 1950s, a group of chemists spent several years assembling a chain of nine amino acids. They got it right, the product behaved like the hormone it was meant to be, and the work was published in 1953 1. It is fair to describe that molecule as one of the most expensive substances ever made, not because the ingredients were precious — they were not — but because the only sensible unit for its cost was a fraction of a scientific career. Nine residues. A Nobel Prize followed.

Seventy years later, the same nine residues can be produced by a plant that will not bother to schedule the run on its own; it will slot the sequence into a campaign alongside forty others. The unit of cost has changed from person-years to a fraction of a currency unit per milligram. What follows is how that happened — four step changes, each smaller than the one before, and one part of the problem that never got cheap at all.

When cost was measured in careers

The pre-1963 method is worth reconstructing, because the shape of its cost explains everything that came after. Peptide bonds were formed in solution. To stop amino acids reacting in every direction at once, every group not currently wanted was capped. Couple, uncap, couple again. That logic was sound, and it is still the logic today. The expense was never in the coupling.

It was in what came next. After each bond formed, the growing chain sat dissolved in a mixture containing unreacted starting material, spent activating agent, its by-products, and fragments that had reacted twice or not at all. All of it had to be separated before the next residue could be added — by crystallisation, extraction, or column chromatography, whichever the intermediate would survive. Each separation consumed a day or more of a trained chemist's time and left product behind on the glassware.

The losses multiplied rather than added. Ten steps recovering eighty per cent apiece leave about eleven per cent of what you started with. Twenty steps leave roughly one. So the cost of a peptide before 1963 was, almost entirely, the cost of skilled human attention multiplied by the number of residues, with a yield penalty compounding against you the whole way. Nothing about that curve suggested it could be flattened.

Who paid for it is part of the answer to why it persisted. Nobody bought peptides in this era, because there was no market to buy them in. They were made inside the institution that wanted them, by people employed to do exactly that, funded by grants and by pharmaceutical research budgets that treated the whole exercise as discovery rather than production. A cost that never appears on an invoice is a cost nobody optimises. The first serious pressure on the price of a peptide arrived only when somebody outside the laboratory that made it wanted to buy one.

The first drop, and the biggest

The 1963 solution was to fix the first residue to an insoluble bead so the chain could not go anywhere 2. Once the product is nailed down, separating it from everything else stops being chemistry and becomes plumbing: open the tap, let the solvent carry the mess away, and the peptide stays behind on the filter. The intermediate purification that had dominated the bill simply disappeared.

Two economic consequences follow, and only the first is obvious. The obvious one is that days of chromatography per residue became minutes of washing. The subtler one is that excess reagent stopped being a liability. In solution work, adding a large excess to force a sluggish coupling made the subsequent separation harder; on a bead, the excess costs one rinse to remove. That inverted the incentive around reagent use and meant couplings could be driven much closer to completion — which, as the arithmetic above shows, is where the real money was.

This was the largest single reduction in the history of the field, and nothing since has come close to matching it. It is also the reason the subsequent steps look modest: once the dominant cost has been removed, everything after it is a refinement of what remains.

What automation actually bought

Because the cycle was now identical every time — deprotect, wash, couple, wash — it could be handed to a machine. The first synthesisers were assemblies of pumps, valves, timers and a shaking vessel, and by the 1970s and 1980s they were commercial instruments sitting in ordinary laboratories.

It is worth being precise about what automation achieved, because the popular version overstates it. Difficult couplings remained difficult. Aggregating sequences still stalled. What automation removed was the requirement that a trained person be present between reactions, and the error rate that comes with fatigue and manual transfers. In cost terms it converted a variable labour expense into a fixed capital expense with a depreciation schedule, which is a smaller change than the bead but a much more predictable one.

The other change of this era was chemical rather than mechanical. The base-labile Fmoc protecting group, introduced in 1972, allowed the temporary protection to be removed under mild basic conditions rather than repeated acid treatment 3. That mattered for cost in an indirect way: milder chemistry meant fewer side reactions, fewer side reactions meant a cleaner crude product, and a cleaner crude product meant less work for the one step that had not been automated at all — the final purification.

The letter soup, and why it mattered

The least visible contribution to the cost curve came from the reagents that activate the carboxyl group so the bond will form at all. This is a genuinely crowded field — a review of it is titled, accurately, more than a letter soup 4 — and its history runs from carbodiimides through active esters to the phosphonium and uronium salts that dominate modern practice.

Their contribution to price is easy to misread, because a better coupling reagent generally costs more per gram than the one it replaces. The saving appears somewhere else entirely. A coupling that succeeds 99.5 times in a hundred rather than 98 produces far fewer chains missing a residue, and chains missing a residue are the single hardest thing to remove from the final product — they are nearly the same size, charge and polarity as the molecule you want. Every fraction of a percentage point of coupling efficiency is paid back at the purification column, which is where the remaining cost lives.

Scale, capacity, and where the floor sits

The fourth step change was industrial rather than scientific. Through the 1990s and 2000s, production moved out of small specialist houses and into large contract plants, most of them in Asia, running campaigns rather than commissions. Reactors got bigger, purification capacity got much bigger, and the fixed costs of a batch — setup, cleaning, analysis, documentation — were spread across far more material.

Something less obvious happened at the same time, and it shows up in the price of short peptides more than anything else. When a plant is sized for pharmaceutical campaigns, the gaps between those campaigns are dead capital, and the cheapest way to fill a gap is to accept small, undemanding work at a price that covers marginal cost rather than full cost. A great deal of the research-grade material in circulation from the late 2000s onwards was priced that way — not as a product with a margin, but as a use for a reactor that would otherwise have been idle. Marginal-cost pricing is a powerful deflationary force, and it is largely invisible from outside the plant.

That is where the decline stops, and it is worth naming what holds the floor up. It is not the amino acids, which are commodity chemicals. It is solvent: the mass of material consumed per unit of peptide produced has been put in the thousands of kilograms per kilogram, far above small-molecule manufacture, and that solvent must be bought, handled under increasingly strict exposure rules, and then destroyed 5. Add preparative chromatography, add lyophilisation, add analysis, add the vial and the cold pack, and you have a cost structure in which the chemistry itself is close to a rounding error.

This piece stops at the factory gate deliberately. What a finished vial sells for anywhere downstream is a question about markets, packaging, testing, regulation and margin, not about synthesis, and the two have been only loosely coupled since about 2005.

The exception that never got cheap

One part of the problem sat out the entire sixty years. Because residues are added one at a time, overall yield is the yield of a single coupling raised to the power of the chain length. At 99 per cent per coupling — a good coupling, not a routine one — a ten-residue peptide finishes around 90 per cent, a fifty-residue peptide just under 61, and a hundred-residue chain near 37. Nothing has gone wrong in any of those figures.

The missing fraction does not vanish; it sits in the same vessel as the product, in the form of chains that skipped a coupling and carried on. Separating a forty-nine-residue impurity from a fifty-residue product is a far harder chromatographic problem than separating a peptide from a solvent, and it gets harder as the chain lengthens. So the cost of a long peptide rises steeply with length while the cost of a short one has been flat for years — which is why chains beyond roughly fifty residues are usually grown in engineered cells or assembled from fragments instead.

Read as a whole, the curve has the shape most manufacturing histories have. One idea removes the dominant cost and everything drops at once. Then decades of incremental work chase what is left, each improvement smaller and harder-won than the last, until the remaining expense is not the clever part of the process but the mundane part — the solvent, the column, the waste contractor. The chemistry stopped being the expensive thing a long time ago. It has taken the rest of the world some time to notice.