The earliest stage of making a medicine is a needle-in-a-haystack problem: out of an almost unimaginable number of possible molecules, which few are worth testing in a lab? That first step is called "hit identification," and according to a News-Medical whitepaper published August 25, 2026, generative AI is starting to change how it is done.
The piece, titled "How generative AI improves hit identification in drug discovery" and also surfaced through Google News and Bing News aggregators, argues that generative AI improves hit identification by accelerating virtual screening and expanding the chemical space researchers can explore.
A short translation for non-chemists. Virtual screening is the practice of testing candidate molecules on a computer before anyone touches a pipette — simulating whether a compound is likely to bind to a disease-related target. "Chemical space" is the full universe of molecules that could in principle exist, which is vastly larger than the libraries of compounds any single company has on its shelves. Generative models can propose new structures rather than only sifting through existing ones, which is what "expanding chemical space" points to.
News-Medical frames these as two of the benefits, indicating there are others it covers in the full whitepaper. The source items available here do not include specific performance numbers, named companies, drug programs, or independent validation, so the claims should be read as an overview of the approach rather than evidence of a particular result.
Why it matters: hit identification sits at the front of a pipeline that routinely takes a decade and billions of dollars, so even modest speedups at this stage ripple through the cost and timeline of every drug that follows.