AI has demonstrated remarkable progress in binding prediction and other stages of drug discovery, but impressive molecules can still fail if not matched to the full translational context. The drug discovery lifecycle spans a chain of interdependent decisions from disease selection to clinical trial strategy. Weak decisions at any point can compromise the program.
As a result, clinical outcomes for AI-discovered drugs remain mixed: many candidates can reach Phase 1, but few have reported convincing human efficacy signals to date. Pancreatic ductal adenocarcinoma (PDAC) illustrates the problem. An AI system might design a potent molecule against validated drivers such as KRAS G12D or KRAS G12V, but potency is only one part of the translation puzzle. PDAC tumors are often desmoplastic and immunosuppressive, limiting drug exposure and making single-agent pathway inhibition vulnerable to adaptive resistance.
This creates an opportunity to approach drug discovery with a vertically integrated platform that spans the full drug discovery lifecycle.
Working backward from the patient, the molecule question broadens into a systems question. An end-to-end platform must identify the optimal patient subgroups, determine whether the drug can reach the tumor, and oversee rigorous preclinical validation. Even the most potent molecules will fail downstream if any of these layers is inadequately addressed.
As outlined in our lab philosophy, critical problems in drug discovery may now be constrained by model judgment, not model intellect. The first plank of our platform builds on this hypothesis by investigating both novel therapies and drug repurposing strategies for rare diseases.
Repurposing is central because it stress-tests translational judgment. In brain tumors, the platform starts from narrow disease biology and CNS delivery constraints, then asks which existing drugs can plausibly reach the tumor, engage the target, and select the right patients. In chordoma, the same logic can evaluate whether a DNA-repair-defined subgroup creates a POLQ/TMEJ dependency, making POLQ inhibition a testable biomarker-gated hypothesis.
These examples capture the larger thesis: the value is not only designing new molecules, but converting known molecules and known biology into testable therapeutic strategies.
