From Algorithm to IV Drip: What AI Drug Discovery Now Means
AI drug discovery is the use of computational models trained on large-scale biological and chemical data to identify novel drug targets and design AI-designed compounds that move into the pharmaceutical development pipeline, aiming to compress timelines from target finding to clinical trial validation while expanding the range of diseases that can be treated. The central shift today is that AI-driven predictions are no longer staying on slide decks; they are being tested in human beings. Recursion’s announcement that its PI3Kα H1047R inhibitor REC-7735, designed with its AI platform, has received IND clearance for a Phase I/II study in select solid tumors is not just a regulatory footnote—it is a line in the sand between hype and evidence. If AI can survive the brutal filter of oncology trials, it will change how risk, cost, and creativity are distributed across drug discovery.
REC-7735: A Quiet but Pivotal Test Case for AI-Designed Compounds
REC-7735’s IND clearance matters because it turns an AI-generated molecule into a potential therapy that must now confront human biology. In practical terms, this is the first major milestone where an AI-designed compound is being judged by regulators and clinicians on equal footing with traditionally discovered drugs. A Phase I/II trial in patients whose tumors carry PIK3CA H1047R mutations is set to begin in the second half of this year, placing REC-7735 squarely into the early-stage oncology ecosystem rather than a niche AI experiment. The message to the industry is blunt: the era of treating AI drug discovery as a side project is over. Once a molecule meets safety and dosing standards for human trials, the debate shifts from whether AI can design “drug-like” structures to whether those structures deliver durable clinical benefit.
Recursion–Genentech: AI Maps That Earn Hard Cash and Optionality
If REC-7735 shows AI can design credible molecules, the Recursion–Genentech collaboration shows AI can discover credible targets. The up-to-USD 12 billion (approx. RM55.2 billion) AI drug discovery alliance reached a key milestone when Genentech exercised its first validated target option for a neuroscience program based on an AI-derived map. That single decision triggers a USD 3 million (approx. RM13.8 million) payment to Recursion and lifts total cash received to USD 216 million (approx. RM993.6 million) since 2021—money that signals confidence more loudly than any conference keynote. More importantly, this target survived a multi-step validation process co-developed with Genentech, including pathway, functional, and disease validation to show that modulating it changes neurological disease phenotypes. In a field where only 8.4% of neurology drug candidates entering Phase I ever reach approval, AI is being asked to do more than accelerate; it is being asked to expand the very map of treatable biology.
Inside the AI Maps: Trillions of Cells, Millions of Images, One Bet on Reproducibility
The neuroscience milestone rests on a radical data-first philosophy. Recursion and its partners built what they describe as the first whole-genome CRISPR knockout map in human neurons, generated from over one trillion internally manufactured neuronal cells derived from induced pluripotent stem cells—about 12 brains’ worth. According to Recursion, candidate targets had to display consistent effects across pathway, functional, and disease validation stages before being advanced into a formal validation package accepted by Genentech. This is not AI freewheeling on synthetic data; it is AI trained on more than 46 million cellular images, each parsed across hundreds of morphological features using proprietary models running on a BioHive-2 supercomputer. The bet here is bold: if you can make biological data generation repeatable at immense scale, AI can discover previously unseen gene–compound and gene–gene interactions and do so in a way that stands up to independent pharmaceutical scrutiny.
From Discovery Platform to Full Pipeline Player
The strategic implication is clear: AI platforms are evolving from service tools into full-stack pipeline engines. Recursion is not stopping at target identification; it plans to use its chemistry platform to design a potential first-in-class molecule for the validated neuroscience target, while continuing to combine its phenomics dataset with Genentech’s transcriptomics data to build additional multimodal maps for new programs. Downstream, REC-7735 moves forward into Phase I/II oncology trials, providing clinical trial validation of AI-designed compounds within a standard development path. Financially, the collaboration structure underscores this shift. For each program, Recursion can earn up to USD 300 million (approx. RM1.38 billion) tied to development, commercialization and net sales milestones, plus tiered royalties in the high single digits for small molecules. AI drug discovery is no longer paid like consulting; it is being paid like a core contributor to pipeline value. The onus now is on AI to deliver not only hits and targets, but approved medicines that justify this new economics.
