Recreating diseases with a patient's induced pluripotent stem cells (iPSCs) and pairing them with artificial intelligence (AI) to assess a drug's efficacy and toxicity in advance could reduce failures in drug development, a speaker proposed.

Joseph Wu, director of Stanford University's Cardiovascular Institute, laid out a people-centered drug development strategy that combines iPSCs, genomic analysis, gene editing and AI. He spoke in a keynote address at the Global Bio Forum 2026, hosted by News1, in the Park Ballroom on the fifth floor of the Conrad Seoul hotel in Yeouido, Seoul, on November 10.

Wu is a cardiologist and physician-scientist whose research uses patient-derived iPSCs to identify the causes of cardiovascular disease and predict how patients will respond to drugs. He is a professor of medicine and radiology at Stanford, served as president of the American Heart Association (AHA) in 2023-2024 and was a member of the U.S. Food and Drug Administration's Cellular, Tissue and Gene Therapies Advisory Committee from 2017 to 2025.

"Each platform is powerful on its own," Wu emphasized. "But they can bring about transformative change only when combined within a framework of human-centered biology."

The key platforms he identified were next-generation sequencing (NGS), CRISPR gene editing, iPSCs and AI. The approach uses the genome to find clues to disease, recreates the disease in patient-derived cells, and then validates drug candidates through AI and experiments.

The strategy centers on iPSCs. These are adult cells, taken from sources such as skin or blood, that have been reprogrammed into stem cells capable of developing into various cell types. Researchers can turn them into cardiomyocytes, or heart muscle cells, or vascular cells, and then study disease patterns and drug responses in cells that carry a patient's genetic traits.

"iPSCs essentially serve as a patient's 'biological twin,'" Wu said.

Using cells that carry a patient's genetic traits can help researchers investigate why people experience different effects and side effects from the same drug. Comparing those cells with ones in which a specific variant has been altered through gene editing can also show how that variant affects disease.

Wu said researchers at Stanford and Greenstone had established more than 2,500 iPSC cell lines that reflect a range of diseases and genetic backgrounds. They are linking the cell lines to medical information and analyzing multiple types of biological data to study how genetic differences affect drug responses.

He also presented examples of recreating disease and drug responses in patient-derived cells. In one study, researchers used cardiomyocytes made from iPSCs of patients with hereditary dilated cardiomyopathy to examine impaired contraction and calcium regulation, and to assess changes after drug treatment.

He also discussed research on the cardiac toxicity of doxorubicin, a drug used to treat breast cancer. The approach compares cells from patients who developed cardiac toxicity after chemotherapy with cells from patients who did not, to examine each person's susceptibility to the toxicity.

The concept of applying such research to drug development is known as the "Clinical Trial in a Dish" (CTiD). Researchers assess a drug's safety and efficacy in patient-derived cells or tissues, then use the results to guide subsequent clinical trials and development decisions.

"This model is particularly promising for rare diseases," Wu said. For diseases with few patients and a wide range of genetic variants, he explained, using patient cells could help determine which drugs should advance to clinical trials.

AI helps identify and prioritize drug candidates. It predicts how likely a compound is to bind to a target protein, as well as its toxicity and other drug properties, narrowing the field of candidates to test.

Wu stressed, however, that experiments must confirm AI's predictions.

"In our experience, many of the drugs predicted by AI do not actually work," he said. "We can confirm that because we synthesize and test the drugs."

He added that researchers use experiments to determine which drugs work and which do not, then feed the results back into the AI to train it further.

The process is iterative: Compounds proposed by AI are tested in a patient's "biological twin," and the results are fed back into the AI. The aim is to improve the accuracy of drug development by continuing to validate predictions in human cells rather than relying on predictions alone.

Wu said new approach methodologies (NAMs), which use iPSCs, organoids and computer models, could help advance preclinical assessments that reflect human characteristics. He outlined a path that would evaluate drug efficacy and toxicity from the earliest stages of development and improve the odds of success in later clinical development.

Recreating diseases with a patient's induced pluripotent stem cells (iPSCs) and pairing th …

How iPSC and AI are Revolutionizing Drug Development: A 2026 Guide