On June 12, 2024, a team of analysts at Morgan Stanley used Anthropic's Claude 3.5 to generate a viable molecular scaffold for a KRAS G12C inhibitor, a target that accounts for roughly 13% of lung adenocarcinomas. The AI‑driven design cut the initial hit‑to‑lead cycle from an industry‑average of six months to just two weeks and identified a compound with an estimated 85% binding affinity improvement over the best publicly disclosed candidate. The experiment was disclosed in a research note that immediately moved the stock of biotech firm Mirati Therapeutics (+7.3%) and sparked a $150 million surge in speculative positions on AI‑enabled drug discovery funds.
The rapid, low‑cost generation of high‑affinity drug candidates demonstrates that large language models can compress a traditionally capital‑intensive R&D phase, potentially reshaping biotech valuation models. If replicated at scale, this could accelerate pipeline turnover across oncology, rare diseases, and beyond, forcing incumbents to integrate generative AI or risk losing competitive advantage.