SP-00 · Before the Pipeline

A windowless conference room, mid-afternoon, the kind of fluorescent light that makes everyone look slightly unwell. Around the table, maybe a dozen people. A few of them are scientists; most of them are not. There is a slide deck, and there is coffee going cold, and there is a decision to be made that none of the people in the room will personally carry out. By the time they leave, they will have committed something on the order of two hundred scientists and the next four years of those scientists’ working lives to a single disease — Alzheimer’s, say, or lupus, or a cancer that strikes nine thousand people a year. The molecules do not exist yet. The targets may not be known. The experiments that will consume those four years have not been imagined, let alone designed.

This is where a drug begins.

It is worth pausing on how strange that is, because almost everyone gets the order wrong. The popular picture of drug discovery starts with a lab — a researcher at a bench, a flash of insight, a compound that turns out to do something remarkable. That picture is not false, exactly. Labs and benches and remarkable compounds are all real, and this curriculum will spend a great deal of time inside them. But they come later. They come after a room like the one above has already decided what the labs will look for and what counts as worth finding. The bench work is downstream of a choice, and the choice is the thing that gets made first.

Call the room a portfolio meeting, because that is roughly what the industry calls it. A pharmaceutical company — or a biotech with one shot and a runway measured in quarters, or a foundation deciding where its donors’ money goes — has more diseases it could attack than it could ever afford to. Every disease it picks is a dozen it has set aside. So someone has to choose, and the choosing happens in advance, on paper, in language that sounds more like strategy than like science: unmet need, competitive landscape, probability of technical success, peak sales. The people who decide are rarely the people who will run the assays. They are deciding what the assays will be for.

If that sounds bloodless — a boardroom carving up human suffering into addressable markets — hold that reaction for a moment, because it is half right and half exactly backwards, and the half that is backwards is the most useful thing this article has to teach. But first, the scale of what is being decided, because the numbers are genuinely difficult to hold in your head, and holding them is the point.

Here is the bet, at the level of the whole industry. Drugmakers worldwide spend on the order of $300 billion a year on research and development — close to $288 billion in 2024, and climbing past $300 billion before the end of the decade. That is the money going in. What comes out the other end, each year, is roughly fifty new medicines — fifty new molecular entities, in the regulator’s phrase, drugs the world has never had before. In 2024 the FDA’s drug-review center approved exactly fifty; in 2023, fifty-five. Fifty a year. It has held there long enough to be a feature, not a fluctuation.

Now do the division, but do it carefully, because it is the single easiest place in this whole subject to mislead yourself. Three hundred billion dollars divided by fifty drugs is about six billion dollars of industry spending per approved drug. That number is true and it is famous and it is constantly, badly misused. It is not the cost of making a drug. No single program costs six billion dollars. The figure is so large precisely because it is an industry-wide ratio — it folds every failure into the price of every success. For each drug that reaches a patient, a great many others died somewhere along the way, in a dish, in a mouse, in a Phase 2 trial that read out flat, and all of that spending is in the numerator. The six billion is what it costs the system to produce one winner, failures and all.

What does a single winning program actually cost? Less, though still a fortune. The most-cited academic estimate, from DiMasi and colleagues, puts the capitalized pre-approval cost — counting the time value of money over a decade-plus of development — at roughly $2.6 billion per approved drug. A more recent estimate from Sertkaya and colleagues, using different accounting, lands lower: about $879 million in expected capitalized cost, or around $172 million if you count only the out-of-pocket cash for the drugs that actually make it. The estimates disagree because they are measuring different things — cash versus capitalized, with failures versus without — and the disagreement is itself worth knowing about (F-12). But a single program in the range of one to three billion dollars is the honest figure to carry. The six-billion number is the industry’s batting average, not the price of a single at-bat. Conflate the two and you will misunderstand every economic argument in this field (F-13).

So: enormous sums, long odds, a handful of survivors. Why does anyone play? Because the medicines market — the money the world spends actually buying and taking drugs, as opposed to the money spent discovering them — is something like $1.6 trillion a year, and growing. (Keep that figure separate in your mind from the R&D spending, and separate again from the much smaller drug-discovery market; they are three different numbers and conflating them is a category error a careful reader catches instantly.) A trillion-and-a-half-dollar market is the gravity well that draws the industry back despite a six-billion-dollar batting average. One Keytruda — Merck’s cancer immunotherapy, which sold roughly $29.5 billion in 2024 alone — pays for an enormous quantity of failure (CS-KEYTRUDA-01).

And the bet is not only large; it is slow. A program chosen in that room today will, if it survives at all, take something on the order of a decade — often ten to fifteen years — to reach a patient. That is the better part of a working career between the decision and the first prescription, which means the room is not really choosing for the world as it is. It is choosing for the world as it will be in twelve years: which diseases will still be unmet then, which competitors will have gotten there first, whose science will have matured in the meantime from impossible to merely difficult. That is a forecast no one can make reliably, and everyone in the room knows it. Part of why so many bets fail is not that the chemistry was wrong but that the question changed while the answer was still being built — a better drug arrived first, or the understanding of the disease shifted under the program’s feet. Time is the quiet third axis of every choice made here, and it is the one no amount of money can buy down.

Which brings us back to the room, and to the assumption I asked you to hold.

The intuitive model of that portfolio meeting goes like this: pharma picks the most lucrative diseases. Follow the money, find the biggest markets, point the scientists there. It is a cynical model and it feels sophisticated, and it is wrong — not because the people in the room are nobler than that, but because the model leaves out the one constraint that actually binds them. The constraint is not which disease would sell. The constraint is which disease can be drugged at all.

This is the misunderstanding worth replacing, so let me replace it carefully. A disease is not a target. A disease is a mess — a tangle of symptoms and mechanisms and afflicted cells, much of it still uncharted. Before anyone can develop a drug, someone has to identify a specific molecular handle inside that mess: a protein, usually, or a gene, that the disease genuinely depends on, and that a drug could plausibly grab hold of and change. That handle is called a target, and the property of having a good one is called tractability — druggable biology with enough evidence behind it to believe that hitting the target will actually move the disease. Tractability is the real gate. Market size only matters for diseases that have made it through. Both target and tractability are among the words the field uses loosely and this series works to use exactly; where a term is carrying more weight than it appears to, the honest glossary is where it is pinned down (X-14).

And many of the largest markets have not made it through. Alzheimer’s afflicts tens of millions and would reward a real treatment beyond almost any other disease — and the field has spent decades and tens of billions of dollars discovering, painfully, how little of its biology is yet tractable; the central mechanism remains genuinely contested, and drug after drug has cleared its presumed target without helping patients much. The market was always enormous. The biology would not cooperate. Meanwhile some far smaller diseases get drugged with comparative speed, because someone found a clean handle — a single broken gene, a single overactive enzyme — that a molecule could reach.

That is why the standard analogy fails and the better one is sharper. The portfolio meeting is not a buyer at an auction, bidding on the most valuable diseases. It is closer to a prospector deciding where to dig — and the prospector cares enormously about where the gold is, but cares first about where the ground can actually be worked. A mountain of gold under bedrock no tool can crack is not an opportunity; it is a tomb for capital. The richest vein you can actually reach beats the richer vein you cannot. The strongest single predictor anyone has found of whether a target will yield an approved drug is not the size of its market but the strength of its biological evidence — targets with human genetic support are roughly twice as likely to make it through (M-SM-04). The room knows this. The room is, above all, hunting for tractable biology with a market attached — and when it cannot find both, biology is the one it cannot fake.

This is the thesis the whole curriculum is built to earn: the pipeline begins with a decision, not a discovery, and that decision is governed first by what biology will permit and only then by what a market will pay. Unmet need tells you a disease is worth treating. Market size tells you a treatment would be worth selling. Tractability tells you whether a treatment is possible at all — and only when all three line up does a program get born (F-16). The 2×2 the strategists draw on the whiteboard has market on one axis and tractability on the other, and the lower-right quadrant — big market, no druggable biology — is where ambitions go to die.

None of this makes the room comfortable. Choosing tractable diseases means, sometimes, declining to work on the most devastating ones, precisely because the most devastating ones are often the least understood. The pancreatic cancers and the rare childhood neurodegenerations and the psychiatric illnesses that wreck the most lives are frequently the ones whose biology offers the fewest handles. A portfolio strategy honest about tractability is, quietly, a strategy that sometimes walks away from the people who need help most — and patient advocacy organizations have learned to fight that logic directly, raising and deploying capital to make an intractable disease tractable enough to enter a room like this one in the first place (X-18). The choice is real, the trade-offs are real, and pretending otherwise would be the kind of smoothing that costs a writer the reader’s trust.

So that is Stage 0: the decision before the science, made by people who will not do the science, constrained first by biology and second by money, at a scale where the system spends six billion dollars for every drug it manages to land. Everything this curriculum covers from here is the working-out of a choice made in that room.

Because once the room has chosen a disease and a target, the questions cascade, and each one is its own stage and its own article. What kind of molecule should attack this target — a small molecule that slips inside the cell, an antibody that patrols the bloodstream (M-MAB-04), a gene therapy that rewrites the instructions outright (M-AAV-05)? How do you find a first candidate, and then make it better? How do you learn whether it is poison before you give it to a person? How do you run the trials that will, against odds of roughly one in seven from the first human dose, either kill the program or prove it — and what do those odds even mean (X-08)? How do you get a regulator to agree, a payer to cover it, a patient to actually receive it?

Every one of those questions is downstream of the windowless room. The scientists at the bench, when they finally arrive, are answering a question that was framed before they walked in — this disease, this target, this much money, this many years. The bet was placed first. The rest of the pipeline is how the bet gets played out.

The lab is where the work happens. The room is where the work is decided. We start, as the industry does, in the room.

The pipeline, stage by stage

Twelve stages take a molecule from a boardroom decision to a medicine on a shelf. Each has its own free installment — read them in order, or jump to the stage you’re curious about.

Stage 1 — The Disease Tells You What to Do — why a drug hunt begins with a molecular target, not a cure, and why most diseases don’t offer one.

Stage 2 — Finding the Knot — how scientists choose the molecule to attack, told through the discovery behind Gleevec.

Stage 3 — Is This Knot Real? — target validation: proving, before the money is spent, that hitting the target actually changes the disease.

Stage 4 — Finding Something That Sticks — hit discovery: screening a million molecules to find the handful worth pursuing.

Stage 5 — Molecular Fine-Tuning — lead optimization: the design–make–test cycle that turns a hit into something drug-like.

Stage 6 — The Last Place You Can Still Be Wrong Cheaply — preclinical testing: the final chance to kill a bad drug before it gets expensive.

Stage 7 — The Application That Lets You Use the Word “Patient” — the IND: how a program earns the right to dose its first human.

Stage 8 — The First Humans — Phase 1: not “is it safe?” but “how much can a body take?”

Stage 9 — Searching for the Signal — Phase 2: where the drug meets real patients, and most fail.

Stage 10 — The Bet — Phase 3: the enormous, decisive trial that measures how well a drug really works.

Stage 11 — The Document That Becomes a Medicine — approval: why the FDA approves a label, not a drug.

Stage 12 — After the Lab — manufacturing and the market: why approval is the starting gun, not the finish.