SP-05 · Molecular Fine-Tuning

A “hit” is a thing of genuine beauty for about a week. Somewhere in a screening library, among hundreds of thousands of molecules, one of them grips the target — the protein you spent the last stage proving was worth attacking (Stage 4 ends here, with that grip confirmed). The assay lights up. The binding is real. For a little while, it is easy to believe you have found a drug.

You have not found a drug. You have found a place to start.

The distance between the two is the subject of this stage, and it is the part of the process that almost nobody outside the lab pictures correctly. The hit binds the target — but it may bind ten other things too, and grip them just as hard. It may dissolve in your test tube but clump into useless gravel in a stomach. It may be potent on a Tuesday and metabolized into nothing by Thursday morning, chewed apart by the liver before it ever reaches the tissue you care about. It may, on a closer look, be quietly mutagenic — the kind of molecule that fixes a cancer by causing three more.

So a small group of people sit in a room and begin to fix these things, one at a time, for years.

The 11pm Whiteboard

On the whiteboard tonight are six molecules — call them analogues, slight variations on last month’s best compound, each differing by an atom here, a ring there. They were designed two weeks ago, synthesized over the following days, and tested this afternoon. The results are in, and they are the usual mixed verdict. Three of the six died in liver microsomes — a standard bench test where you expose a compound to the enzymes that do most of the body’s metabolic demolition, and watch how fast it disappears. These three disappeared fast. Two more are insoluble; they will not dissolve in anything resembling blood. One of them is almost there: potent, reasonably stable, modestly soluble. Almost.

Tomorrow the chemists will design the next six, reasoning from what these six taught them. Then the cycle runs again. And again.

This is the loop at the center of the whole enterprise, and it has a name that flatters it slightly: the design–make–test cycle. You design a molecule on a hypothesis, make it (which means a synthetic chemist actually builds it, bond by bond — see F-04 for what that involves), and test it against a battery of assays. The data refute your hypothesis or refine it, and you design again. This is lead optimization, Stage 5 of the pipeline, and it is the least cinematic and most consequential work in drug discovery. Here is the thesis the whiteboard is teaching: a drug is not discovered in a flash of binding. It is optimized into existence, slowly, against many constraints that fight each other — and the central difficulty is that improving any one property tends to wreck another.

What the Hit Is Missing

It helps to name what a drug actually has to do, because “binds the target” is only the first item on a long list.

A real drug has to be potent — it has to work at a dose small enough to be practical, which usually means binding the target tightly. It has to be selective — it has to bind that target and not the dozen structurally similar proteins doing essential jobs elsewhere, because every off-target grip is a potential side effect. It has to be soluble enough to dissolve and travel in the watery medium of the body. It has to survive that body long enough to act: to get absorbed (often from the gut, if it is a pill), to distribute to the right tissue, to resist being metabolized into junk too quickly, and eventually to be excreted cleanly. Pharmacologists bundle that survival story under the acronym ADME — absorption, distribution, metabolism, excretion — and the time-course of a drug’s rise and fall in the blood is its pharmacokinetics, or PK (the curves that govern all of this are the subject of F-05).

Now the cruel part. These properties are not independent dials you can each turn up. They are coupled, and often they pull in opposite directions. The chemical features that make a molecule grip its target tightly — greasy, flat, aromatic surfaces that nestle into a protein’s binding pocket — are frequently the same features that make it refuse to dissolve in water. So you add a polar, water-loving group to fix solubility, and your potency drops because the molecule no longer fits the pocket as snugly. You tune the shape back toward potency, and now it slips into an off-target protein and your selectivity collapses. You block that off-target binding, and the new chemical group you added to do it turns out to be metabolized in an hour.

This is the multi-objective squeeze, and it is the real texture of the work. You are not solving one problem. You are solving five or six at once, where every solution is a new problem, and the win condition is a single molecule that is merely good enough at all of them simultaneously. That is rarer than being excellent at any one.

A Worked Example: How Imatinib Got Built

The cleanest illustration of this squeeze is the molecule that became imatinib — sold as Gleevec, the drug that turned a lethal leukemia into a managed condition. Its full origin story has its own installment (the compound very nearly was never made at all — see CS-GLEEVEC-02); here we care only about the chemistry of how a crude starting compound was tuned into a medicine.

The chemists at Ciba-Geigy began with a lead from a class called the phenylaminopyrimidines — molecules that inhibited kinases, the signaling enzymes whose malfunction drives the leukemia. The lead was a starting point in exactly the sense this article means: interesting, not a drug. What followed was a sequence of deliberate modifications, each one a worked instance of the squeeze.

First, potency. They attached a particular chemical group — a 3’-pyridyl group — at a specific position, and the compound’s ability to shut down the target inside living cells jumped. Good. But the molecule was still promiscuous: it also inhibited protein kinase C, an enzyme the body needs, which is precisely the selectivity problem that turns a drug into a liability.

So, selectivity. They added what the chemists nicknamed a “flag-methyl” group — a single small methyl group hung off the anilino ring like a flag. That one addition abolished the unwanted activity against protein kinase C and bought them the selectivity they needed. A tiny change; a large consequence.

Now the molecule was potent and selective. It was also nearly useless as a pill, because it would not dissolve. So, solubility. They appended a polar side chain — an N-methylpiperazine group — that markedly improved how well the compound dissolved in water and, with it, how much survived the trip from gut to bloodstream when swallowed. This is the textbook move: bolt on a water-loving group to rescue solubility.

And here the squeeze closes its jaws. Adding that side chain to the existing structure left an aniline group exposed in a configuration that carried mutagenic potential — the molecule now risked damaging DNA, which is disqualifying. To fix that, the chemists had to insert an amide linker and a benzene-ring spacer into the molecule, restructuring it to neutralize the hazard the solubility fix had created.

Read that sequence again, because it is the whole lesson in miniature. Potency fix creates a selectivity problem. Selectivity fix holds. Solubility fix creates a mutagenicity problem. Mutagenicity fix requires rebuilding part of the scaffold. Every move that solved one thing opened another. The finished molecule — the one that became Gleevec — is not the product of a single insight. It is the accumulated answer to a long argument between constraints, each atom present because some earlier version failed without it.

SAR: Reading the Structure–Activity Map

The reason chemists can navigate that argument at all — rather than mutating molecules at random — is a discipline called structure–activity relationship, or SAR. The idea is simple and powerful: you change one piece of the molecule, you measure how the activity changes, and you build up a map of which structural features matter and which don’t. Move the methyl group here, potency doubles. Move it there, potency vanishes. Swap this ring for that one, selectivity improves but solubility craters. Each synthesized analogue is a data point; the map that emerges tells the chemist where to push next.

SAR is why the imatinib story reads as a sequence of reasoned moves rather than lucky accidents. The flag-methyl was not a guess. It was a hypothesis — this position controls the off-target binding — tested by making the molecule and measuring the result. This is also where the modern tools are reshaping the oldest craft in the building: machine-learning models now help predict which analogues are worth making before anyone synthesizes them, compressing the design half of the cycle (see X-16), and structure-prediction tools let chemists reason about the target’s shape with a fidelity that was science fiction a decade ago (see X-02). The loop is the same. The design step is getting smarter.

It is tempting, faced with all this trade-off juggling, to want a checklist — a set of rules that tell you whether a molecule is “drug-like” before you bother. There is a famous one, and its fate is instructive.

The Rule of Five, and Why Good Drugs Break It

In 1997, a chemist named Christopher Lipinski and his colleagues at Pfizer looked across a large set of orally successful drugs and noticed a pattern. The ones that worked as pills tended to stay within certain bounds: not too heavy, not too greasy, not too many of the particular chemical groups that donate or accept hydrogen bonds. He codified it as the “rule of five” — the limits cluster around the number five and its multiples — and it became the most cited heuristic in medicinal chemistry. A molecule that broke several of the rules was, the thinking went, unlikely to make a good oral drug.

It is a genuinely useful rule. It is also broken constantly, and increasingly on purpose.

Of all the FDA-approved small-molecule drugs across history, only about 51% are both taken orally and fully compliant with the rule of five. And that share has been falling: among new medications approved between 2013 and 2019, roughly 40% violated at least one of the rules. Whole therapeutic classes live cheerfully outside it — many antibiotics, antifungals, vitamins, and the cardiac glycosides that have steadied failing hearts for two centuries. A more recent generation of “beyond-rule-of-five” drugs, such as the cancer drug venetoclax, are large, complicated molecules made orally viable not by obeying the rule but by clever formulation and chemical tricks that escort them into the body anyway.

The point is not that Lipinski was wrong. The point is that lead optimization is trade-offs, not a checklist. The rule of five is a useful prior — a sense of where the easy territory lies — and a good chemist knows precisely when to spend effort leaving that territory because the target demands it. Treating the heuristic as a law would have killed venetoclax in a spreadsheet. The constraints are real, but they are constraints to be negotiated, not boxes to be ticked.

The Phrase to Unlearn: “A Drug Hits Its Target”

There is one mental image, almost universal, that quietly distorts everything in this stage, and it is worth dismantling on its own.

People say a drug “hits its target,” and the verb does the damage. Hit sounds like a switch being flipped — the drug arrives, the target is engaged, the job is done, on becomes off. It is the picture of a key turning in a lock once and staying turned. Almost nothing about that is right, and the wrongness matters for how you think about everything that follows.

Here is the truer picture. A drug and its target are not locked together permanently; they are in constant negotiation. Molecules of the drug bind to molecules of the target, and then let go, and then bind again — an equilibrium of grabbing and releasing happening billions of times over. At any given instant, the drug occupies only a fraction of the available target molecules. That fraction — call it the occupancy — is set by how much drug is present and by how fast the molecule grabs on and lets go. And because the amount of drug in your blood rises after a dose and falls as the body clears it, the occupancy is not even a fixed fraction. It is a curve that climbs and recedes over hours.

So the right replacement for “a drug hits its target” is this: a drug raises and sustains the occupancy of its target over time. You are not flipping a switch. You are turning a dimmer, and then watching how long the light stays up. This reframing — receptor occupancy as the genuinely useful way to think about drug action, with all its consequences — is important enough that it gets its own full treatment later (F-15). For now, hold the corrected image, because it explains the squeeze. When a chemist optimizes for potency, they are partly tuning how tightly and how long the drug stays bound — the residence time. When they optimize for PK, they are shaping the curve of how much drug is present to do the binding. Potency and pharmacokinetics are not two separate goals. They are two handles on the same occupancy curve, which is why pulling one so often moves the other.

How Long, and How Many

It is fair to ask what this costs in sheer effort, and the honest answer is a range, because the literature counts differently depending on where it draws the boundaries.

During lead optimization proper, a program might synthesize somewhere on the order of 500 to 1,200 compounds — every one designed, built, and tested through the cycle on the whiteboard. Count the earlier hit-to-lead work too, and the figure can climb toward several thousand molecules made in pursuit of a single candidate. Pull back to the whole discovery funnel and the attrition is starker still: on the order of 5,000 to 10,000 molecules are screened and worked on for every one that eventually becomes an approved drug — roughly one or two survivors per ten thousand. These numbers are not precise constants; treat them as the shape of the thing, not a measurement. The shape is what matters. This is a grind measured in years of iteration and many, many cycles, most of which end with a molecule on the whiteboard that almost worked.

And this is only one modality’s version of the grind. What “optimization” means changes with the kind of drug. For a small molecule, it is the atom-by-atom SAR sculpting described here (the small-molecule path is M-SM-05). For a therapeutic antibody, optimization looks like affinity maturation — refining the protein’s binding region through cycles that echo how the immune system itself improves antibodies (M-MAB-05). For a PROTAC or molecular glue, you are optimizing a molecule that must grip two proteins at once and bring them together, a fundamentally different geometry problem (M-PROTAC-05). For an antisense oligonucleotide, the optimization happens in the chemistry of the genetic backbone itself (M-ASO-05). Same loop, different sculpture.

The Bench Is Not the Body

At the end of all this, the whiteboard has a winner. After years of cycles, one molecule is potent enough, selective enough, soluble enough, stable enough, and clean enough — a compound that is, at last, good at everything at once. The chemists have something they are willing to call a candidate.

But notice what that judgment rests on. Every property was measured on a bench — in assays, in microsomes, in cells, in dishes. The liver-microsome test approximates metabolism; it is not a liver. The solubility measurement is in a buffer; it is not a bloodstream after a meal. A molecule that looks good on the bench has cleared the conditions a laboratory can simulate, and the body is far stranger and less forgiving than any simulation. It has organs that concentrate drugs in unexpected places, immune systems that take offense, individual variation that no assay anticipates.

Which is exactly where the next stage begins. We have a candidate now — a molecule built, atom by argued-over atom, to be a drug. Before it can be given to a single human being, it has to prove it can survive a living system and not poison it. That is the work of preclinical development, and it is where many beautiful molecules quietly die.

This is the free spine of The Lead Compound.

You’ve just read one stage of how a medicine actually gets made. The spine is the free through-line — the whole pipeline, start to finish. The full course goes deeper: every drug class (antibodies, mRNA, cell and gene therapy, peptides, and more) and the real, documented stories behind the medicines that defined them — Ozempic, Keytruda, Gleevec, Comirnaty, and dozens more.

← Previous (SP-04) · How to read this course · Next: (SP-06) →