Appearance
25.3 — The Branches: Pharma, Biotech, Microbiology, Devices, Diagnostics
Aspirin weighs 180 daltons. A dalton is the unit chemists use for the mass of a single molecule, roughly the mass of one hydrogen atom, so aspirin is a molecule of about twenty-one atoms. You can draw it on a napkin.
Trastuzumab, a breast cancer antibody, weighs about 148,000 daltons. It is a folded protein of over a thousand amino acids, made inside living hamster cells in a steel tank, and no chemist can synthesise it. Nobody can even fully describe it: two batches are never atom-for-atom identical, and the rules accept that.
Those two products need different factories, different scientists, different quality tests, different shipping, different regulators inside the same agency, and different law when somebody wants to copy them. That difference is what the word "branches" means in this industry, and getting it wrong in a meeting is the fastest way to look new.
The small-molecule pharmaceutical business
A small molecule is a chemical compound of low molecular weight — normally under 900 daltons — made by chemical synthesis and usually swallowed as a tablet or capsule. Aspirin, paracetamol, atorvastatin, metformin, amoxicillin, sertraline: almost everything in your bathroom cabinet.
The properties that follow from being small are the reason this branch behaves the way it does.
It survives the gut, so it can be a tablet, which is why small molecules dominate long-term daily treatment. It can cross cell membranes, so it can act on targets inside the cell, which biologics generally cannot reach. It is stable, so it sits on a shelf at room temperature for years, and the supply chain is ordinary freight rather than refrigerated freight. It is cheap to make once discovered — the pill costing a patient a dollar may cost cents to produce, because the expense was the fifteen years of discovery and trials, not the manufacturing.
And crucially, it can be copied exactly. Another company can synthesise the identical molecule, prove it reaches the bloodstream the same way, and sell it as a generic. Chapter 25.15 works through how that proof is done.
The organisations here are the ones whose names you know, plus a huge and mostly invisible layer beneath them: the manufacturers of APIs, meaning active pharmaceutical ingredients, the pure drug substance before it becomes a tablet. A large share of the world's API supply is made in India and China, which is why drug shortages are usually supply-chain events rather than scientific ones.
The biotechnology business
Biotechnology, in the drug sense, means using living cells as the factory. The product is a large biological molecule — a protein — that living cells make because they were given the gene for it.
The families you will meet are these.
Therapeutic proteins that replace something a person lacks — insulin, growth hormone, clotting factors for haemophilia, erythropoietin for anaemia.
Monoclonal antibodies, which are the largest category by value. An antibody is the immune system's targeting molecule, described in Chapter 13.3, and this branch makes vast quantities of one single antibody chosen to stick to one chosen target: a receptor on a cancer cell, an inflammatory signal in rheumatoid arthritis, a cholesterol-regulating protein in the liver.
Vaccines, which train the immune system rather than treating a disease directly, covered mechanically in Chapter 13.5.
Cell and gene therapies, where the product is living cells or a piece of DNA or RNA meant to enter your cells and stay. These are the newest and hardest, and Chapter 25.4 takes them properly.
Everything difficult about this branch comes from the fact that the product is a large, folded, fragile molecule made by cells.
It cannot be swallowed, because your digestive system exists to break proteins into fragments (Chapter 9.3), so nearly all of these are injected or infused. It is fragile, so it usually needs refrigeration from the factory to the patient, and a temperature excursion in transit can destroy a shipment worth millions. It can be attacked by the patient's immune system, because it is a foreign protein; this is called immunogenicity and it is tested for specifically. And the process is the product — change the cell line, the growth medium, or the purification steps, and the molecules that come out are subtly different. That is the single most important sentence in biologics manufacturing, and Chapter 25.11 shows what it forces a factory to do.
Where microbiology sits, and why it is everywhere
People are often surprised that microbiology is listed as a branch of the industry rather than a subject in a textbook. It is here because microbes appear in four completely different roles, and confusing them causes real errors.
Microbes as the enemy. Antibiotics, antifungals and antivirals are developed against them (Chapters 17.1 to 17.3 and 22.6), and resistance testing is a permanent laboratory function.
Microbes as the factory. Bacteria and yeast are the cheapest cells to grow, so they make insulin, many enzymes and many vaccine components. Fermentation science is a career of its own.
Microbes as the contaminant, which is the one that governs the factory. Any injected product must be sterile, meaning free of living organisms. A sterile manufacturing plant runs constant environmental monitoring — settle plates, air samplers, surface swabs, personnel gowning checks — and every one of those results is a record somebody must review. A large share of the deviations discussed in Chapter 25.17 begin as a microbiology result that fell outside its limit.
Microbes as the patient's own ecosystem. The gut microbiome (Chapter 9.6) is now a target for products in its own right.
The medical device business
A device is anything that acts on the body by physical means rather than by chemistry inside the body. The legal definitions differ slightly by country, but that plain test is nearly always right: if the main action is mechanical, electrical, thermal or optical, it is a device; if it works by pharmacological, immunological or metabolic action, it is a drug.
The range is enormous and it is easy to underestimate. A wooden tongue depressor is a device. So is a syringe, a catheter, a hip replacement, a hearing aid, a pacemaker, a dialysis machine, an MRI scanner, an infusion pump, and a surgical robot.
And so is software, if it makes a medical claim. An application that reads a scan and reports a suspected bleed is a medical device in its own right — software as a medical device, usually shortened to SaMD. This is the doorway through which most software engineers enter regulated territory without realising it, and Chapter 25.16 covers the rulebook that then applies, including the standard for the software lifecycle itself.
The economics of devices differ sharply from drugs and this shapes the projects. Development is usually cheaper and faster, iterations are frequent, the product may be revised annually, and it is sold to hospitals and surgeons rather than prescribed to patients. The failure mode is different too: a bad drug harms whoever swallowed it, while a bad implanted device may harm everyone who received it and require surgery to remove. That is why device regulation leans heavily on traceability — knowing exactly which unit went into which person, through the UDI, the unique device identifier.
The diagnostics business
Diagnostics do not treat anything. They produce information, and the industry treats them as a branch because the information changes what happens to the patient.
In-vitro diagnostics, or IVD, are tests performed on a sample taken out of the body: blood chemistry, cultures, PCR tests, pregnancy tests, the analysers in a hospital laboratory. Imaging is the other half: X-ray, ultrasound, CT and MRI machines, explained physically in Chapter 16.6.
Two ideas from this branch keep appearing in software projects.
The companion diagnostic. Some modern drugs only work in patients whose tumour carries a specific molecular change, so the drug is approved together with the test that finds it. Trastuzumab is the classic case: it is for tumours that overproduce the HER2 protein, and the test decides who gets the drug. A companion diagnostic is approved as a device, tied to the drug's label, and Chapter 25.4 explains why this is the practical meaning of "precision medicine".
And laboratory regulation is separate from device regulation. In the United States, the laboratory that runs the test is regulated under CLIA, the Clinical Laboratory Improvement Amendments, which govern how a lab must be run and how its staff must be qualified — a completely different rulebook from the one covering the instrument it runs the test on.
Which rulebook applies to your project
This table is the practical payoff of the whole chapter. When a client describes a product, work out which row it is in, because the row determines the regulator, the approval path, and the standards your software must satisfy.
| Product | US pathway | Copy is called |
|---|---|---|
| Small-molecule drug | NDA | Generic |
| Biologic, incl. antibodies | BLA | Biosimilar |
| Vaccine | BLA | Rarely copied |
| Cell or gene therapy | BLA | Not yet routine |
| Most moderate-risk devices | 510(k) | Not applicable |
| High-risk devices | PMA | Not applicable |
| Diagnostic test kit | Device pathway | Not applicable |
NDA is a new drug application, BLA a biologics license application, 510(k) a notification that a device is essentially the same as one already on the market, and PMA a premarket approval, the full evidence route for the highest-risk devices. Chapters 25.13 and 25.16 take each apart properly; for now the value is in recognising which world you are in.
One more distinction worth carrying, because it explains organisation charts. Inside the US regulator, drugs and biologics are reviewed by different centres — CDER for drugs and most therapeutic antibodies, CBER for vaccines, blood products and cell and gene therapies, and CDRH for devices and diagnostics. A client's regulatory team is usually organised the same way, and asking "which centre is this going to?" is a question that marks you as someone who has been here before.
The service and supplier layer underneath all of it
None of the branches above is self-sufficient, and the outsourced layer is where most service work actually lands.
CRO — contract research organisation, runs trials. CDMO — contract development and manufacturing organisation, makes product. CMO is the older term for the manufacturing-only version. Central labs analyse trial samples. Site management organisations run networks of trial sites. Specialty pharmacies handle complex or expensive medicines that ordinary pharmacies do not stock. Wholesalers and distributors move nearly everything.
And then the information layer: electronic data capture vendors, safety database vendors, regulatory publishing vendors, laboratory information systems, hospital record systems, claims processors, analytics companies. Your company is somewhere in that last sentence, and the branch your client belongs to decides which regulations follow your code into production.
Next: Chapter 25.4, what is actually happening right now — cell and gene therapy, mRNA platforms, and an honest accounting of what artificial intelligence has and has not yet done to drug discovery.