Most drugs in human history were discovered by accident, observation, or trial and error long before anyone understood why they worked. Ancient civilizations chewed bark, brewed leaves, and applied plant extracts based on centuries of accumulated experience. Over time, chemistry, biology, and technology transformed drug discovery from folklore into a systematic process, though serendipity still plays a surprising role. The path from a promising compound to an approved medicine now takes over a decade and costs hundreds of millions of dollars.
Plants, Bark, and Traditional Medicine
For thousands of years, the primary method of drug discovery was simple: people noticed that certain plants made them feel better, and that knowledge was passed down through generations. Willow bark was used to treat pain and fever as far back as ancient Egypt and Greece. Chemical investigation into why it worked didn’t begin until the early 1800s, when researchers started isolating the active substance. In 1828, a chemist at the University of Munich removed the impurities and obtained a yellowish compound he called salicin, after the Latin name for willow. Further refinements eventually produced salicylic acid and then acetylsalicylic acid, which we know today as aspirin.
This pattern repeated across cultures and continents. Quinine, used to treat malaria, came from the bark of the cinchona tree long used by indigenous peoples in South America. Artemisinin, another antimalarial, was identified in the plant Artemisia annua, which had been part of traditional Chinese medicine for centuries. The cancer drug paclitaxel (sold as Taxol) was isolated from the bark of the Pacific yew tree. Vinblastine and vincristine, two chemotherapy drugs still in wide use, come from the Madagascar periwinkle. In each case, centuries of traditional knowledge pointed scientists toward the right plant, and modern chemistry did the rest.
Unmodified natural products, compounds used exactly as they appear in nature, now represent roughly 5% of all FDA-approved drugs. The most common classes are alkaloids, certain peptides, polyphenols, and polyketides. Approvals of these unmodified natural drugs have declined by about two per decade since the mid-1900s, largely because raw natural compounds often need chemical modification to work safely and reliably as medicines. But nature still provides the starting scaffolds that chemists then refine.
The Role of Accidents and Side Effects
Some of the most important drugs in history were discovered entirely by chance. In 1928, Alexander Fleming noticed that a mold had contaminated one of his bacterial cultures in a London laboratory, and the bacteria near the mold were dying. He isolated the mold, identified it as belonging to the Penicillium genus, and named its active agent penicillin. He published his findings in 1929, but then nothing happened for a full decade. No one could figure out how to produce enough of the substance to use it as a medicine.
That changed in 1939, when Howard Florey assembled a team at Oxford that included Norman Heatley, a fungal expert, and Ernst Chain, a biochemist who successfully purified penicillin from the mold extract. They tested it on mice infected with deadly bacteria: four treated mice survived, four untreated mice died. The first human to receive penicillin was an Oxford policeman with severe infections, treated in February 1941. By September 1943, production had scaled enough to supply the Allied armed forces during World War II. The strain eventually used for mass production wasn’t even Fleming’s original mold. It came from a moldy cantaloupe found at a market, and it produced six times more penicillin than Fleming’s strain.
Accidental discoveries didn’t stop in the 20th century. In 1986, Pfizer scientists were developing a compound to treat angina, the chest pain caused by reduced blood flow to the heart. During clinical trials, the drug didn’t do much for angina, but male volunteers kept reporting an unexpected side effect: penile erections. Pfizer pivoted, ran new trials, and the drug became sildenafil, sold as Viagra. It later found a third life as a treatment for pulmonary hypertension. This kind of repurposing, where a drug designed for one condition turns out to work for something else entirely, remains a productive path in modern medicine.
Designing Drugs on Purpose
By the late 20th century, advances in molecular biology made it possible to design drugs with a specific target in mind rather than stumbling onto them. This approach, called rational drug design, starts with understanding the precise molecular machinery that goes wrong in a disease, then building a compound to interfere with it.
One of the most celebrated examples is the leukemia drug imatinib, sold as Gleevec. Chronic myelogenous leukemia is caused by a mutation that leaves a specific enzyme permanently switched on, driving uncontrolled cell growth. Researchers studied the three-dimensional structure of that enzyme and designed a molecule that fits into a particular groove on its surface, locking it into an inactive shape. The drug was remarkably selective: it hit its intended target powerfully while largely sparing similar enzymes in healthy cells. Gleevec transformed a once-fatal cancer into a manageable chronic condition and became a proof of concept that drugs could be engineered with precision rather than found by luck.
Screening Millions of Compounds
Not every drug can be designed from scratch. Much of modern drug discovery relies on high-throughput screening, an industrial-scale approach where robotic systems test vast libraries of chemical compounds against a biological target to see which ones have any effect. The numbers are staggering: roughly one million compounds must be screened to produce a single marketable drug. Most candidates fail at every stage, from initial lab testing through animal studies and human clinical trials.
The process is slow and expensive. Developing a new drug costs an estimated $879 million on average when you factor in the cost of all the failed candidates along the way and the time value of the money invested. Some estimates run much higher depending on the therapeutic area, with pain and anesthesia drugs among the most expensive to develop. The nonclinical stage alone, before any human testing begins, typically takes over two and a half years. Total timelines from early discovery to approval commonly stretch beyond a decade.
Genetic Engineering and Biologics
A major turning point came when scientists learned to use living cells as drug factories. In 1978, researchers at Genentech produced the first recombinant human insulin by inserting the human insulin gene into E. coli bacteria, which then manufactured the protein. By 1982, this lab-grown insulin reached the market, replacing insulin harvested from pig and cow pancreases. It was the first pharmaceutical product made using recombinant DNA technology.
This opened the door to an entire class of medicines called biologics: large, complex molecules produced by living organisms rather than synthesized in a chemistry lab. Biologics now include monoclonal antibodies used in cancer treatment, immune-modulating drugs for autoimmune diseases, and gene therapies that correct genetic defects at their source. They represent one of the fastest-growing segments of modern medicine, and their discovery process looks fundamentally different from traditional small-molecule chemistry.
How AI Is Changing the Process
The newest shift in drug discovery involves artificial intelligence. AI systems can analyze molecular structures, predict how compounds will behave in the body, and suggest new drug candidates far faster than human researchers working through trial and error. The greatest impact so far has been in the earliest stages of discovery, where AI significantly reduces the time needed to identify promising molecules and optimize their properties. A process that once took years of lab work can now be compressed into months of computational modeling, though much of this evidence still comes from proof-of-concept studies rather than fully approved drugs.
AI doesn’t replace the need for animal testing and human clinical trials, which remain the longest and most expensive parts of the pipeline. But by narrowing the field of candidates earlier and more accurately, it has the potential to reduce the enormous failure rate that makes drug development so costly. Several AI-discovered compounds have already entered human trials, and the pharmaceutical industry is investing heavily in the technology as a way to bring down both timelines and costs.