Intermediateapplied biology
Pharmaceutical Biotechnology: How Modern Drugs Are Made
Learning objectives
- Outline the drug discovery pipeline from target to approval.
- Compare biologics with small-molecule drugs in design and manufacturing.
- Describe what happens in each clinical trial phase.
- Explain pharmacovigilance and why drugs are sometimes withdrawn.
- Assess AI's real versus hyped contributions to drug discovery.
The Pipeline: From Target to Drug
Drug discovery starts with a target, usually a protein whose activity drives disease, identified through genetics, disease biology, or basic research. High-throughput screening then tests hundreds of thousands of compounds against the target, or rational design builds molecules to fit its structure. Lead compounds undergo optimization: medicinal chemists tweak the structure for potency, selectivity, and drug-like properties such as solubility and metabolic stability. Animal studies assess toxicity and behavior in living systems before any human dosing. The sobering statistic frames everything: roughly nine of ten candidates entering clinical trials fail, most for lack of efficacy or safety problems, which is why the successes must fund the failures.
Small Molecules Versus Biologics
Traditional drugs are small molecules, a few dozen atoms synthesized by chemistry, taken as pills that distribute throughout the body. Biologics are large proteins, antibodies, hormones, enzymes, manufactured by engineered cells. Small molecules can enter cells and cross the blood-brain barrier but often hit unintended targets, causing side effects. Biologics bind their targets with exquisite specificity but are confined mostly to extracellular spaces, require injection, and need cold-chain storage. Manufacturing differs utterly: chemistry scales predictably, while living-cell production is finicky and expensive. Modern pipelines use both: small molecules for intracellular targets and brain diseases, biologics for immune disorders, cancers, and conditions needing precise protein replacement.
Clinical Trials in Detail
Phase 1 enrolls dozens of volunteers to establish safety, dosing, and how the body handles the drug. Phase 2 expands to hundreds of patients, testing whether the drug actually works against the disease and refining the dose. Phase 3 randomizes thousands against the current standard of care, measuring whether the new drug improves meaningful outcomes like survival or symptom control. Each phase has predefined endpoints and statistical thresholds to prevent wishful interpretation. Independent data-monitoring boards can halt trials early for overwhelming benefit or emerging harm. Only after all three phases do regulators weigh the complete evidence; approval is a judgment that benefits outweigh risks for specific patients, not a declaration of universal safety.
After Approval: Vigilance and Withdrawal
Approval is not the end of testing. Phase 4 pharmacovigilance monitors millions of users for rare side effects invisible in trials of thousands, through adverse-event reporting systems and observational studies. Drugs are sometimes restricted or withdrawn when real-world data reveal unacceptable risks: rofecoxib (Vioxx) was pulled after cardiovascular risks emerged in large-scale use. Manufacturing problems can also force withdrawals. This ongoing surveillance is a feature, not a failure: no trial can detect one-in-fifty-thousand harms, so the system is designed to keep learning after launch. Patients contribute directly by reporting suspected side effects, making pharmacovigilance a genuine partnership between regulators, companies, clinicians, and the public. Patients contribute directly to this safety net by reporting suspected side effects, making pharmacovigilance a genuine partnership rather than a bureaucratic afterthought.
AI in Drug Discovery: Signal and Noise
Artificial intelligence genuinely accelerates parts of discovery: predicting protein structures, screening virtual compound libraries, and optimizing molecules faster than traditional methods. AlphaFold's structure predictions shortcut years of crystallography, and generative models propose novel chemical scaffolds. But the honest accounting notes what AI has not solved: biology's complexity means most AI-designed candidates still fail in trials, because predicting human efficacy remains far harder than predicting binding. Several AI-discovered drugs have entered clinical testing, a real milestone, yet none of the industry's throughput problems have vanished. The realistic view treats AI as a powerful tool that compresses the early pipeline while the expensive, slow biology of human trials remains the bottleneck.
AI speeds molecule design, but proving safety and efficacy in humans still requires years of trials. Computation cannot shortcut clinical evidence.
Key points
- Drug discovery moves from target identification through screening, optimization, and animal testing.
- Small molecules are cheap pills with broad distribution; biologics are precise but complex living-cell products.
- Phase 1 tests safety, Phase 2 tests efficacy signals, Phase 3 proves benefit in thousands.
- Post-approval pharmacovigilance catches rare harms and sometimes forces withdrawal.
- AI accelerates early discovery but has not removed the clinical-trial bottleneck.
Common misconceptions
✓ Truth: Animal studies filter out many toxic candidates but predict human responses imperfectly; species differences in metabolism and biology explain why drugs still fail in human trials after passing animal tests.
✓ Truth: Safety depends on dose, chemistry, and evidence, not origin. Many natural substances are toxic, and unregulated products lack the purity and interaction testing required of pharmaceuticals.
✓ Truth: Testing escalates cautiously: dozens in Phase 1, hundreds in Phase 2, thousands only in Phase 3 after earlier phases justify it. Each step requires safety data from the last.
Frequently asked questions
Why are new drugs so expensive?
Because failures fund successes: about 90 percent of candidates fail after costly testing, development spans a decade or more, and manufacturing biologics is inherently expensive. Prices also reflect patent-protected market exclusivity rather than production cost alone.
What happens in each trial phase?
Phase 1 establishes safety and dosing in dozens of people; Phase 2 tests early efficacy in hundreds; Phase 3 proves benefit against standard care in thousands. Regulators review the complete package before approval.
Why do approved drugs get withdrawn?
Usually because post-approval surveillance in millions of users reveals rare but serious harms invisible in trials of thousands, as with Vioxx's cardiovascular risks. Withdrawal shows the safety system working after launch.
How is AI actually changing drug discovery?
It accelerates target identification, structure prediction, virtual screening, and molecule optimization, compressing years of early work into months. But AI-designed candidates still face the same clinical trial gauntlet, and human efficacy remains unpredictable.