- Atlas
- Medicine & Health
- Drug Discovery
Drug Discovery
Drug discovery is the work of finding molecules that can safely change the course of a disease, and of proving that they do.
13 frameworks, from 50 to today · Pharmacology · Medicine & Health
- Materia medica 50-1850
- Active-principle isolation 1805-1930
- Receptor theory 1900-Present
- Magic-bullet chemotherapy 1909-1945
- Systematic empirical screening 1935-1990
- Rational drug design 1950-Present
- Structure–activity relationship 1950-Present
- Pharmacogenomics 1955-Present
- ADME 1960-Present
- Combinatorial chemistry and high-throughput screening 1990-2010
- Fragment-based lead discovery 2000-Present
- Genomics-driven target discovery 2001-Present
- Computational and AI-driven drug discovery 2010-Present
Overview
Drug discovery is the work of finding molecules that can safely change the course of a disease, and of proving that they do. It is older than any of its component sciences: people were testing plants, minerals, and poisons on the sick long before anyone could say what a molecule was. What has changed across its history is not the goal but the method—how a discoverer decides where to look, and how much can be known before the first patient is dosed. That shift, from collecting remedies to designing molecules against a chosen biological target, is the spine of the story.
Materia medica: remedies collected and catalogued
For most of recorded history, drug discovery meant materia medica—the accumulated body of substances used as medicines, with their preparations, doses, and reputed effects. The Greek physician Dioscorides compiled a De materia medica in the first century CE that described some six hundred plants and other substances and remained a working reference for centuries1. Chinese, Indian, Islamic, and European traditions each built their own pharmacopoeias, and the knowledge travelled: opium, cinchona bark, digitalis, and willow bark entered practice through such collections, often with little understanding of why they worked.
The materia medica approach had a real strength—it preserved hard-won empirical observations—and a structural weakness: it could not distinguish an active substance from the bulk of a plant, could not standardize a dose, and could not explain an effect. Those limits set the agenda for the next two centuries.
Isolating the active principle
The first great turning point was chemical. If a plant worked, the reasoning went, some specific compound in it must be responsible. In the early nineteenth century, chemists learned to extract and purify such compounds: morphine was isolated from opium by Friedrich Sertürner around 1804–1806, and quinine from cinchona bark by Pierre-Joseph Pelletier and Joseph Caventou in 1820. Alkaloid chemistry followed, and with it the idea of a pure drug whose dose could be measured in milligrams rather than in handfuls of leaf.
Isolation made standardization possible and made pharmacology a laboratory science, but it also exposed a new problem: knowing a compound's structure did not reveal what it did in the body, or to which tissue.
Receptors, targets, and the magic bullet
Around 1900, Paul Ehrlich proposed that drugs act not on organs in general but on specific chemical structures in cells—receptors—and coined the term magic bullet for a drug that would strike a disease-causing agent while sparing the patient. Ehrlich's own work on arsenicals against syphilis, culminating in Salvarsan (arsphenamine) in 1909, was the first great demonstration that a synthetic chemical could be aimed at a defined target23. In parallel, John Newport Langley's experiments on nicotine and curare suggested that drugs acted at specific receptive sites, and later A. J. Clark's quantitative work in the 1920s and 1930s formalized the idea that a drug's effect depends on how much of a receptor it occupies.
Receptor theory gave drug discovery its modern question—what is the target, and how tightly does the drug bind it?—and made the search for new drugs a search for molecules with the right shape and affinity.
Systematic screening and the rise of pharmacokinetics
From the 1930s onward, discovery increasingly meant screening: testing many compounds against a biological assay and keeping the ones that work. This was systematic empirical screening, and it produced a remarkable run of drugs—sulfonamides, penicillin and the antibiotic era, antihypertensives, and many others. The approach did not require knowing the target; it required a reliable assay and a large collection of compounds.
A second strand matured alongside it: pharmacokinetics, the study of what the body does to a drug—absorption, distribution, metabolism, and excretion, often abbreviated ADME4. By the 1960s and 1970s it was clear that a compound could be potent in a test tube and useless in a patient because it was poorly absorbed, rapidly destroyed by the liver, or unable to reach its target. ADME optimization became a routine part of discovery, and later computational and in vitro screens were built to predict these properties before a compound was ever given to an animal.
Rational design and structure–activity relationships
The next turning point was the deliberate use of molecular structure. Structure–activity relationship (SAR) work—systematically varying a molecule's chemical groups and measuring how potency changes—became a standard tool, and it allowed chemists to improve a lead compound step by step56. The antihistamine and beta-blocker programs of the mid-twentieth century were built this way.
Rational drug design went further: start from knowledge of the target's three-dimensional structure and design a molecule to fit it7. The classic early case is the work of Gertrude Elion and George Hitchings at Burroughs Wellcome in the 1950s, who reasoned from the biochemistry of nucleic acid synthesis to design antimetabolites—6-mercaptopurine and later azathioprine and allopurinol—that interfered with specific enzymatic steps. Captopril, approved in 1981, was designed from knowledge of the angiotensin-converting enzyme. Rational design made the target, not the compound library, the starting point.
High-throughput screening and combinatorial chemistry
In the 1990s, automation and chemistry converged. Combinatorial chemistry allowed chemists to synthesize vast libraries of related compounds at once, and high-throughput screening (HTS) allowed robotic assays to test hundreds of thousands of compounds per run. The promise was that sheer scale would find hits faster than design could. In practice, the era produced fewer new drugs than hoped—libraries were often poorly suited to drug-like chemistry, and hits frequently failed later—but it permanently changed the infrastructure of discovery and pushed the field toward better libraries and better assays.
Fragments, genomics, and the return to structure
Two responses to the limits of HTS shaped the 2000s. Fragment-based drug discovery started from very small molecules that bind weakly to a target and then grew or linked them into potent drugs; it proved especially valuable for targets that resisted conventional screening. Separately, the sequencing of the human genome, announced in 2001, promised a flood of new targets—genomics-driven target discovery. The promise was real but slower than expected: a gene identifies a protein, not necessarily a good drug target, and validating a target takes years.
Pharmacogenomics: the patient enters the picture
A parallel line of work asked why the same drug helps one patient and harms another. Pharmacogenomics—the study of how genetic variation affects drug response—grew from mid-twentieth-century observations of inherited differences in drug metabolism, such as the differing rates at which people acetylate isoniazid, and matured with molecular genetics. It gave discovery a new consideration: a drug might be developed alongside a genetic test that identifies the patients likely to benefit, the companion diagnostic. This moved the field from "does the drug work?" toward "in whom does it work?"
Computational and AI-driven discovery
The most recent turn is computational. Machine learning models now predict binding, toxicity, and ADME properties, and generative methods propose new molecular structures; the field calls this AI-driven drug discovery. The approach descends directly from earlier computational chemistry and from the data generated by HTS and structural biology. Its promise is to shorten the search; its open question is whether predictions made in silico survive the transition to real assays and real patients.
Where the field stands
Today's discovery pipelines combine most of these strands: a target chosen from genetics or disease biology, screening and fragment methods to find starting points, structure-based design and SAR to optimize potency, ADME and toxicity prediction to filter candidates, and increasingly pharmacogenomic thinking to define the right patient population5. The unresolved questions are the old ones in new form—how to validate a target before investing years in it, how to predict human toxicity from animal and cellular data, and how to make computational predictions trustworthy enough to replace experiments. The history of drug discovery is, in the end, a history of learning to ask sharper questions of molecules before trusting them with patients.
References
- De materia medica, Wikipedia
- Magic bullet (medicine), Wikipedia
- Paul Ehrlich, Wikipedia
- ADME, Wikipedia
- Drug discovery (Screening and design, introduction), Wikipedia
- Structure–activity relationship, Wikipedia
- Drug design (Rational drug discovery, Structure-based), Wikipedia
The frameworks of Drug Discovery
Each framework is a way of working the field has used: when it held, and the question it answers.
- 50-1850
- 1805-1930
- 1900-Present
- 1909-1945
Magic-bullet chemotherapy
How can a drug hit the germ and spare the patient?
- 1935-1990
Systematic empirical screening
Which compound works in a biological test?
- 1950-Present
- 1950-Present
Structure–activity relationship
How does changing a molecule change its potency?
- 1955-Present
- 1960-Present
- 1990-2010
Combinatorial chemistry and high-throughput screening
Can testing vast numbers of compounds find drugs faster?
- 2000-Present
Fragment-based lead discovery
Can tiny weakly binding pieces be grown into drugs?
- 2001-Present
Genomics-driven target discovery
Which genes point to good drug targets?
- 2010-Present
Computational and AI-driven drug discovery
Can computers predict which molecules will work?
How the ideas connect
Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.
Active-principle isolation grew out of Materia medica
Active-principle isolation grew directly out of materia medica: the observation that certain plant preparations worked prompted chemists to find the specific compound responsible. Sertürner's isolation of morphine from opium around 1805 and Pelletier and Caventou's quinine in 1820 turned the collected remedy into a purified substance.
Carried by Friedrich Sertürner; Pierre-Joseph Pelletier; Joseph Caventou.
Rational drug design influenced by Receptor theory
Rational design starts from a chosen target, and the target concept is receptor theory's. Elion and Hitchings designed antimetabolites against specific enzymatic steps, treating the enzyme as the receptive site to be fitted and blocked.
Carried by Gertrude Elion; George Hitchings.
Combinatorial chemistry and high-throughput screening grew out of Systematic empirical screening
High-throughput screening is systematic empirical screening automated and scaled up: the same assay-and-keep-what-works logic, run robotically on hundreds of thousands of compounds. Combinatorial chemistry supplied the large libraries that HTS needed.
Receptor theory influenced by Active-principle isolation
Once pure compounds could be tested, the question became how they act on the body. Ehrlich's receptor concept and Langley's receptive sites were framed around purified drugs and poisons whose doses could be measured, a practice made possible by active-principle isolation.
Carried by Paul Ehrlich; John Newport Langley.
Structure–activity relationship influenced by Active-principle isolation
Structure–activity work required pure compounds of known structure, which active-principle isolation and alkaloid chemistry had made available. Varying chemical groups and measuring potency presupposed the purified, characterized molecule.
Systematic empirical screening influenced by Receptor theory
Receptor theory supplied the biological assays that systematic screening depended on: a measurable response at a receptive site gave screeners a readout for whether a compound worked. Clark's quantitative occupancy work in the 1920s and 1930s made such assays interpretable.
Carried by A. J. Clark.
Systematic empirical screening influenced by Magic-bullet chemotherapy
The magic-bullet aim of striking a disease agent while sparing the patient carried into screening programs, which sought selective antibacterials and antiparasitics. The sulfonamide and antibiotic era pursued exactly that selectivity through empirical testing.
ADME influenced by Systematic empirical screening
Systematic screening produced potent compounds that failed in patients, which forced attention onto absorption, distribution, metabolism and excretion. The repeated finding that a good assay result did not predict a good drug drove ADME into routine discovery practice.
Fragment-based lead discovery influenced by Rational drug design
Fragment-based lead discovery is structure-based design applied to small starting pieces: fragments are grown or linked using knowledge of the target's three-dimensional structure. It extends the rational-design premise that the target structure, not the library, is the starting point.
Combinatorial chemistry and high-throughput screening reacted against Structure–activity relationship
Combinatorial chemistry was promoted as an alternative to slow, one-compound-at-a-time SAR optimization, promising that sheer library scale would find hits faster than design. In practice the libraries were often poorly suited to drug-like chemistry, and the field returned to SAR-guided optimization.
Computational and AI-driven drug discovery influenced by ADME
ADME prediction was an early computational success and became a standard machine-learning task in AI-driven discovery. Models that predict absorption, metabolism and toxicity in silico carry forward the ADME question of whether a compound will survive in the body.
Fragment-based lead discovery reacted against Combinatorial chemistry and high-throughput screening
Fragment-based discovery arose as a response to the limits of HTS: instead of screening large drug-like libraries, it started from very small weakly binding molecules and grew them. It was framed as a way to reach targets that conventional high-throughput screening missed.
Computational and AI-driven drug discovery influenced by Combinatorial chemistry and high-throughput screening
AI-driven discovery descends from the data HTS generated: the large screening datasets became training material for machine-learning models predicting binding and properties. Computational chemistry methods built on that data were extended by generative models.
Computational and AI-driven drug discovery influenced by Fragment-based lead discovery
Fragment-based discovery contributed structural and binding data that computational methods use to propose and grow molecules. Generative design extends the fragment-growing logic with machine-learned models rather than manual structure inspection.
Beyond the field
Where Drug Discovery's ideas came from in other fields, and what they led to there: the strongest 12 of 96 links.
Receptor theory influenced Receptor-Targeted Imaging in Molecular Imaging
Receptor theory supplied the concept of specific high-affinity ligand binding to receptors that receptor-targeted imaging exploits to design tracers. The imaging approach is the application of receptor-binding principles to visualize receptor distribution in vivo.
Receptor theory influenced Rational drug design in Pharmacology
Receptor theory gave rational drug design its founding premise that drugs act at specific molecular sites, so structures can be designed to fit a known target.
Magic-bullet chemotherapy influenced Secondary Metabolite and Antibiotic Discovery in Mycology
The magic-bullet ideal of a drug that harms the parasite but not the host is the explicit aim that antibiotic discovery pursued with penicillin and streptomycin. Ehrlich's programme set the target that the antibiotic era realised with microbial products.
Carried by Paul Ehrlich; Alexander Fleming.
Magic-bullet chemotherapy influenced Chemotherapy in Medical Oncology
Magic-bullet chemotherapy of 1909-1945, built on Ehrlich's arsenicals and the sulphonamides, gave anticancer chemotherapy its founding aim of selective cytotoxicity.
Carried by Paul Ehrlich.
Rational drug design influenced by Receptor theory in Pharmacology
Receptor theory gave rational drug design its central premise that a drug acts at a specific receptor whose structure can be modelled and targeted.
Rational drug design influenced by Receptor theory in Medicinal Chemistry
Receptor theory as used in medicinal chemistry is the conceptual basis for designing ligands to fit a characterised receptor.
Rational drug design led to Structure-Based Drug Design in Medicinal Chemistry
Structure-based drug design is the branch of rational drug design that uses the three-dimensional structure of the molecular target, growing directly out of the earlier receptor-fitting programme.
Rational drug design led to Mechanism-Based Inhibitor Design in Medicinal Chemistry
Rational drug design, built on knowing the target and designing molecules to fit it, is the parent programme from which mechanism-based inhibitor design grew: the inhibitor is designed around a known enzymatic mechanism rather than a static fit.
Rational drug design led to Computational and In Silico Drug Design in Medicinal Chemistry
Computational drug design grew out of rational drug design as its computational branch: the receptor-fitting and physicochemical-property reasoning of rational design was implemented in software for docking, QSAR and de novo design.
Structure–activity relationship grew out of Medicinal chemistry in Pharmacy
Structure–activity relationship analysis is a core method of medicinal chemistry, which grew from nineteenth-century pharmacy and chemistry into the systematic design and optimization of biologically active compounds.
Pharmacogenomics grew out of Pharmacology in Pharmacy
Pharmacogenomics grew out of pharmacology as a branch studying how genetic differences modify drug action, following observations of inherited drug-response variants such as suxamethonium sensitivity and primaquine haemolysis.
Combinatorial chemistry and high-throughput screening influenced Combinatorial and high-throughput synthesis in Materials Chemistry
Combinatorial chemistry and high-throughput screening, developed in drug discovery in the 1990s, supplied combinatorial and high-throughput materials synthesis with its core methods: split-pool libraries, parallel synthesis and automated screening. The materials field explicitly transferred these techniques from pharmaceutical discovery.

Drug Discovery is a planet in Nousfarer
Its city has a tower for each of these 13 frameworks. Land, walk in, and climb one concept at a time.
Land on Drug DiscoveryRelated fields
Fields its ideas are linked to
- Medicinal Chemistry 21 links
- Pharmacology 11 links
- Pharmaceutical Chemistry 7 links
- Pharmacy 5 links
- Microbiology 4 links
- Structural Biology 3 links
- Computational Chemistry 3 links
- Genomics 3 links
- Pharmacodynamics 3 links
- Pharmaceutics 3 links
- Ethnobotany 3 links
- Environmental Toxicology 2 links