The gap between technological promise and clinical reality defines one of the most consequential frontiers in modern medicine. For decades, drug development has suffered from attrition rates exceeding 90%, timelines stretching 10–15 years, and costs surpassing $2 billion per approved compound. Artificial intelligence now offers systematic tools to compress these bottlenecks — but a comprehensive review in Critical Reviews in Oncology/Hematology clarifies precisely where that promise is being realized and where it remains aspirational.
The review catalogues how generative AI models — including scaffold-aware design platforms, 3D molecular architecture tools such as AlphaFold, MoleR, and PocketCrafter, and large language models — are accelerating target identification, lead optimization, and phenotypic screening across the drug development pipeline. Single-cell foundation models are extending AI's reach into precision biology, enabling more granular characterization of disease mechanisms. Critically, the authors document that while the FDA has approved numerous AI-enabled medical devices and software tools, no fully AI-conceived and AI-designed drug has yet received marketing authorization, even as several AI-originated candidates have advanced into clinical trials.
This review arrives at a pivotal inflection point. The field is moving from proof-of-concept demonstrations toward genuine clinical translation, but the review's most analytically important contribution is its articulation of where AI's utility hits hard ceilings: data quality upstream and experimental validation constraints downstream. Generative models can explore vast chemical spaces, but they remain only as good as the biological and chemical databases they train on — many of which contain significant noise, bias, and gaps. The review implicitly challenges the prevalent narrative that AI will dramatically shorten development timelines in the near term, suggesting instead that current gains are concentrated in early-stage discovery rather than clinical execution. For health-conscious adults, the practical implication is measured optimism: AI is meaningfully improving the odds that better-targeted, less toxic therapies eventually reach patients, particularly in oncology — but the biology still controls the timeline.