Drug Development: Targets, Trials, and the Pipeline

The Headline Numbers

Rough figures for a new drug, 2026 vintage:

Time from target identification to approval:     10 to 15 years
Cost (all-in, counting failures):                $1 to $2.5 billion
Attrition rate (targets to approval):            ~1 in 10,000 compounds
Revenue potential (blockbuster):                 $1+ billion/year
Patent life remaining at approval:               typically 10 to 12 years

The economics are brutal. Most drug candidates fail. Those that succeed often have only a few years of exclusive sales before patents expire. The successes pay for the failures.

This shapes everything about the biotech industry: why it's expensive, why drugs cost what they do, why companies pick the targets they pick.

The Pipeline

The standard development pipeline:

Target identification        understand what protein or pathway drives a disease
Target validation            confirm modulating it helps
Hit finding                  find molecules that affect the target
Lead optimisation            improve those molecules
Preclinical                  test in cells and animals
Phase 1 trial                safety in healthy volunteers (usually)
Phase 2 trial                dose-finding and early efficacy
Phase 3 trial                large-scale confirmatory efficacy and safety
Regulatory approval          FDA, EMA, or other
Post-market                  continued monitoring

Each step has high failure rates. Most programs die in preclinical or Phase 2.

Target Identification and Validation

Every drug program starts with a target: a protein or pathway whose modulation addresses a disease.

How targets are found:

• Genetic studies (variants causing disease point to proteins)
• Disease biology research
• Chemical biology (a compound works; find out why)
• Phenotypic screens (look for compounds that fix a disease model, then figure out what they hit)

Target validation: does modulating this target actually help? Methods:

• Animal models with the gene knocked out
• Human genetic studies (people with loss-of-function variants as natural experiments)
• Cell-based assays
• Early biomarker studies

Many targets look promising and turn out to be wrong. A drug that modulates a target that doesn't actually drive the disease won't work in patients, no matter how well the drug works biochemically.

Hit Finding

Once a target is set, you need molecules that affect it. Two main strategies:

High-throughput screening (HTS)

Test a library of compounds (often 100,000 to a few million) against the target and see what hits. Industrial-scale automation. Gets you a starting point, usually a weak hit that needs improvement.

Structure-based drug design

Solve the structure of the target, identify the binding pocket, design molecules that fit. Requires a tractable target structure; works best when you have co-crystal structures with existing ligands.

DNA-encoded library (DEL) screening

Test very large libraries (billions of compounds) in a single tube using DNA tags to track which hit. Newer; produces more diverse hits.

AI-assisted design

Models trained on known ligands and structures propose new molecules. Growing in importance; not a replacement for experimental validation.

Lead Optimisation

A hit is usually not a drug. It binds the target but has problems: weak affinity, poor solubility, gets metabolised too fast, toxic, can't cross membranes, doesn't reach the right tissue.

Lead optimisation iterates: modify the molecule, test, repeat. A team of medicinal chemists adjusts the structure over months or years to hit all the requirements simultaneously.

Properties they optimise:

• Potency: how tightly it binds the target
• Selectivity: how little it binds the wrong targets
• Pharmacokinetics (PK): how it moves through the body (absorption, distribution, metabolism, excretion)
• Safety: no cardiac, liver, or other toxicity
• Formulation: can you make a pill, an injection, an inhaler?

Optimisation ends when the molecule passes enough hurdles to move to preclinical testing. Often that's a year or more of chemistry.

Preclinical

Testing in cells and animals before humans.

• In vitro: cell cultures, biochemical assays
• In vivo (animals): rodents first, then often larger species
• Toxicology: what are the safety signals? Reproductive effects? Cardiac? Liver? Carcinogenic?

This stage generates the IND (Investigational New Drug) application in the US. The FDA reviews; if they accept, you can start human trials.

Clinical Trials

Phase 1

First in humans. Usually 20 to 100 healthy volunteers (for most drugs; oncology often uses patients). Goals:

• Safety: what doses are tolerated?
• Pharmacokinetics: how does the drug behave in humans?
• Maximum tolerated dose: the dose at which side effects become unacceptable

Phase 1 takes a few months to a year. Most drugs clear Phase 1 (around 60-70%); Phase 1 failure usually means unexpected toxicity.

Phase 2

Patients with the target disease, typically 100 to 500 of them. Goals:

• Dose finding: what dose works, with acceptable side effects
• Initial efficacy: does the drug actually help the disease?
• Continued safety

Phase 2 is where most drugs die. Roughly 30% make it through. "Didn't work" in Phase 2 often means the target was wrong, or the drug doesn't reach the target at a safe dose.

Phase 3

Large confirmatory trials, 300 to many thousands of patients. Usually randomised, often placebo-controlled (or compared against standard of care). Goals:

• Demonstrated efficacy at statistically significant scale
• Safety profile in a larger population
• Comparison against existing treatments

Phase 3 takes years and costs hundreds of millions of dollars. About 60% of Phase 3 trials succeed.

Approval

Regulatory bodies (FDA in the US, EMA in Europe, and others) review the data and decide. Approval usually comes with labels specifying approved uses and warnings.

Phase 4 / Post-Market

After approval, continued monitoring for rare side effects, long-term effects, and expanded indications. Some drugs are withdrawn years after approval when problems emerge (Vioxx is a famous case).

Attrition Math

A rough overall success rate:

Preclinical candidates → approved drugs:   ~1 in 10,000
Phase 1 → approval:                        ~10%
Phase 2 → approval:                        ~30%
Phase 3 → approval:                        ~60%

Most things die. The survivors pay for the deaths.

Drug Modalities

Different kinds of drugs have different properties:

Small molecules

Traditional pills. Easy to manufacture, cross cell membranes, orally bioavailable. Limited in what they can target (mostly proteins with well-defined binding pockets).

Biologics (proteins, antibodies)

Larger molecules: monoclonal antibodies, fusion proteins, cytokines. Need injection. Very specific, often very expensive. Revolutionised oncology and autoimmune disease treatment.

Cell therapies

Living cells as therapy. CAR-T (chimeric antigen receptor T cells) for leukaemia and some lymphomas. Take the patient's T cells, engineer them to attack their cancer, return them. Complex manufacturing, spectacular results in some cases.

Gene therapies

Delivering DNA or modified DNA to correct a disease. AAV-based therapies for retinal disease (Luxturna), spinal muscular atrophy (Zolgensma). High cost; some are million-dollar one-time treatments.

mRNA

Deliver mRNA encoding a protein; cells make it. Used in COVID vaccines and now in development for cancer vaccines, rare disease therapies.

Antisense oligonucleotides

Short RNA-like molecules that bind target mRNAs and modulate their processing or degradation. Nusinersen for SMA; eteplirsen for Duchenne muscular dystrophy.

Each modality has its own economics, manufacturing complexity, and disease applicability.

Who Pays and Why Drugs Cost What They Cost

Drug pricing is a political, ethical, and economic issue. A simplified version:

• R&D is expensive: the industry argues prices recoup development costs
• Patent protection: approved drugs have limited exclusive sales windows
• Risk: for every approved drug, many failures were funded
• Price-sensitivity varies: oncology patients and rare-disease patients pay more (because they have to) than chronic-disease patients
• Insurance intermediation: in the US, insurance negotiates prices; list prices are rarely what's paid
• International variation: other countries have price controls; the US largely doesn't

Critics argue prices are higher than justified by R&D costs (marketing is huge; profits are high). Industry argues pricing supports future innovation. Both positions have merit; neither is the whole story.

Biotech vs Big Pharma

The industry has two kinds of companies, roughly:

Biotech

Smaller, younger, often focused on one or a few programs. Higher risk, higher reward. Most biotech companies fail; those that don't are often acquired by big pharma.

Big Pharma

Large incumbents (Pfizer, Merck, Roche, AstraZeneca, Novartis, etc.). Many marketed drugs, diversified pipelines. Acquire biotechs with promising programs to fill their own pipelines.

The division of labour: biotech innovates (often out of academic discoveries), pharma commercialises. Messy in practice; many pharmas have serious research; many biotechs have commercialisation capabilities.

The Valley of Death

A persistent problem: the gap between academic discovery and an investable biotech company.

• Academics make a discovery with clinical potential
• Venture capital wants further de-risking before funding
• Traditional biotech funding is hard to access at very early stages
• Many promising discoveries languish

Efforts to bridge: accelerator programs, translational research funding, venture builders. Progress is uneven; the valley is still real.

Biotech and AI

AI is reshaping multiple parts of the pipeline:

• Target identification: ML on multi-omics data to find new targets
• Drug design: generative models proposing novel molecules
• Protein structure: AlphaFold enabling targets previously inaccessible
• Clinical trial design: better patient selection, synthetic controls
• Patient stratification: finding who responds to which drug

AI doesn't shortcut biology; biology still happens in wet-lab and in patients. AI speeds parts of the loop. Whether it reduces overall timelines and costs substantially is still being determined.

Common Pitfalls

"Company X's drug showed promise in trials." Phase of trial matters enormously. Phase 1 "promise" is weak signal; Phase 3 "promise" is strong. Read carefully

"The drug works." Works in what population? For which outcome? At what dose? Compared to what? All of these matter

"Approval is the finish line." Approval is the gate; market uptake, reimbursement, and real-world effectiveness are the next race. Many approved drugs fail commercially

"Big Pharma is evil / is saintly." Oversimplifications. The industry has produced enormous medical value; it has also priced and marketed in predatory ways. Both are true

"AI will dramatically shorten drug development." Has sped some steps. Hasn't yet shortened approval timelines in aggregate. Cautious optimism is warranted; revolutionary claims less so

Next Steps

Continue to 12-best-practices.md for the habits that keep you reading biotech well after this tutorial ends.