The best-selling drugs in the world were never designed. They were selected, the way evolution selects

For most of pharmaceutical history a drug was a small molecule that a chemist could draw, synthesise and modify. The best-selling medicines in the world are now mostly not that. They are monoclonal antibodies — large proteins, hundreds of times the size of an aspirin molecule, grown in vats of animal cells rather than synthesised.

The interesting thing about them is not their size. It is that nobody designs them. A chemist improving a small molecule reasons about which group to move where. An antibody against a new target is not reasoned about at all: you set up a situation where billions of candidates exist, apply a pressure that only the right ones survive, and see what comes out. The design work is outsourced — to an immune system, or to a selection run in a test tube. Only afterwards does human engineering begin.

The molecule, briefly

An antibody is a Y. The two tips of the Y carry loops of highly variable sequence, and those loops fold into a surface that grips one specific patch on one specific molecule. The stem is constant and does a completely different job: it is the handle the rest of the immune system holds, telling it what to do with whatever has been grabbed.

The body generates that variety by shuffling gene segments during the development of each B cell, so that every B cell ends up producing one antibody of one specificity. The immune system is, in effect, a standing library of billions of binders waiting for something to match. A monoclonal antibody is the product of a single such cell, cloned — one molecular sequence, one target, made identically forever. That is the difference from the antivenom-style preparation raised in an animal's blood, which is a mixture from many cells at once.

Two ways to find the one you want

Hybridoma was the original method and it is beautifully practical. Immunise a mouse with the target. Its immune system does exactly what it evolved to do and produces B cells making antibodies against that target — but B cells die quickly in culture. So you fuse them with myeloma cells, which are cancerous and immortal, and the resulting hybrid inherits both properties: it makes one antibody and it divides forever. Then you separate the hybrids, grow each one, and test them one by one until you find the clone making the antibody you wanted. Köhler and Milstein published this in 1975 and it won a Nobel Prize within a decade.

Phage display removes the animal entirely and is closer to laboratory evolution. Build a library of billions of antibody fragments, each one genetically fused to a coat protein of a bacteriophage so that the virus particle carries the binder on its surface and the gene encoding it inside. Pour the library over your immobilised target, wash away everything that does not stick, elute what does, and grow it up in bacteria. The binders are now enriched. Repeat three or four times and a library of billions collapses to a handful of strong binders, whose sequences you can read directly because each phage carries its own blueprint. The method was recognised with a Nobel Prize in 2018, shared with directed evolution — which is exactly what it is.

An antibody drug is not drawn, it is fished for. Almost all the cleverness sits in designing the selection so that only the molecules you want survive it.

Editing the mouse out

Early therapeutic antibodies came from mice, and patients' immune systems noticed. A mouse protein injected into a human is foreign, so the body raises antibodies against the drug itself. The treatment stops working, and reactions follow.

Fixing this drove twenty years of protein engineering. Chimeric antibodies kept the mouse binding tips and replaced the rest with human sequence. Humanised antibodies went further, grafting only the six short hypervariable loops onto an otherwise entirely human framework. Fully human antibodies come from phage libraries built on human sequence, or from transgenic mice whose own antibody genes have been replaced with human ones.

This history is written into the older drug names, which is a satisfying thing to be able to read. A name ending in -omab is mouse-derived; -ximab is chimeric; -zumab is humanised; -umab is fully human. That is how rituximab, trastuzumab and adalimumab announce their construction in their last four letters. The naming system was revised in 2021 and newer approvals use different stems, so the trick only works on the older generation — but that generation includes most of the antibodies you have heard of.

What the stem is for

The binding end gets the attention; the constant stem is where much of the modern engineering happens.

It determines effector function — whether the antibody merely blocks its target or actively recruits immune cells to destroy whatever it has attached to. For a cancer antibody you usually want that recruitment; for an antibody blocking an inflammatory signal you want it silenced, because killing the cells displaying that signal is not the goal. The stem can be engineered either way.

It also explains a property that makes these drugs practical. Antibodies bind a recycling receptor inside cells that rescues them from degradation and returns them to the bloodstream. That salvage pathway gives a typical therapeutic antibody a half-life of around three weeks, which is why a biologic can be injected fortnightly or monthly while a small molecule must be taken daily. Engineering the stem to bind that receptor more tightly extends the interval further.

Using the antibody as a chassis

Once you can reliably make a protein that finds one thing in the body, the molecule becomes a delivery system.

An antibody–drug conjugate attaches a cytotoxic payload to the antibody through a chemical linker designed to hold in the bloodstream and release inside the target cell. The payload is typically far too toxic to administer on its own; tethering it to something that only docks at tumour cells is what makes it usable. Most of the difficulty is in the linker, which must be simultaneously stable and cleavable at the right moment.

A bispecific antibody is engineered with two different arms. The most striking use grips a tumour cell with one arm and a T cell with the other, physically dragging an immune cell into contact with a target it was ignoring — a chemical version of the problem CAR-T solves by re-engineering the T cell itself.

And there are smaller formats. Camelids make antibodies with a single binding domain, and these nanobodies are a fraction of the size, penetrate tissue better and tolerate conditions that would destroy a conventional antibody.

Why they cost what they do

Antibodies cannot be made in bacteria. They are large, they must be assembled from multiple chains, and they carry sugar structures added by the host cell that materially affect both effector function and clearance — a bacterium attaches none of them. So production runs in mammalian cell culture, usually Chinese hamster ovary cells, in bioreactors held for weeks under tight control, followed by a purification train with its own capture and polishing steps.

That is a fundamentally more expensive process than fermentation of a simple protein, and it is the main reason antibody therapies are priced where they are. It is also why biosimilar versions require their own comparability evidence rather than a simple equivalence test — the same constraint that governs every protein drug.

Why it matters for students and researchers

The conceptual lesson is the one worth carrying: for a whole class of modern medicine, the discovery step is a selection rather than a design. Set up the library, set up the pressure, and let it find the answer. That idea now runs well beyond antibodies — into enzyme engineering, aptamers, peptide discovery and the machine-learning methods that are beginning to propose binders computationally rather than fish for them.

For India the practical relevance is direct. The country has a substantial biosimilar antibody industry and growing mammalian cell-culture capacity, and the constraint there is rarely the science — it is people who can run and characterise a mammalian process to regulatory standard: cell line development, glycan analysis, bioreactor control, comparability studies. Those are unglamorous, specific and genuinely scarce skills, and they sit between the biology and the factory rather than in either one.

Frequently asked questions

What is a monoclonal antibody?

It is an antibody produced from a single cloned cell line, so every molecule is identical and binds the same specific target. This distinguishes it from polyclonal preparations, such as antivenom, which contain many different antibodies from many cells.

How are monoclonal antibodies discovered?

Either by immunising an animal and immortalising its antibody-producing cells through hybridoma fusion, or by building a library of billions of antibody fragments displayed on bacteriophage and selecting the ones that bind the target through repeated rounds of washing and amplification.

Why were early antibody drugs modified to be more human?

Because mouse-derived proteins provoke an immune response in patients, which neutralises the drug and can cause reactions. Chimeric, humanised and fully human antibodies progressively replaced mouse sequence, a history recorded in the older drug-name endings -ximab, -zumab and -umab.

What is an antibody-drug conjugate?

It is an antibody carrying a toxic payload attached by a chemical linker. The antibody delivers the payload selectively to cells displaying its target, allowing the use of compounds far too toxic to give on their own.

Why are antibody drugs expensive?

They must be produced in mammalian cell culture rather than bacteria, because they are large multi-chain proteins carrying sugar modifications that affect how they work. That process requires long bioreactor runs, extensive purification and tight quality control.