Cofolding Advances Erode Pearl's Lead

Pearl’s edge matters only if it stays unique long enough to change how drugs are selected in real programs. The market is moving too fast for benchmark wins alone to hold that position. AlphaFold 3 already pushed protein and small molecule joint structure prediction into the mainstream, Lilly has started distributing internal small molecule and antibody models through Benchling workflows, and Isomorphic Labs is building a full stack around structure, binding, properties, and toxicity. That shifts the contest from best model demo to best data, best workflow, and fastest path to a real drug candidate.

• Genesis says Pearl beats AlphaFold 3 and other open models on public cofolding benchmarks, including OpenBind. That is important because cofolding predicts the 3D pose of a protein and a candidate drug together, which is the step chemists use to decide what molecule to synthesize next. But once several groups can do this well enough, the bottleneck moves downstream to wet lab proof.
  1 sacra 6 sacra
• Large incumbents are no longer waiting for outside platforms to mature. Lilly launched TuneLab in January 2025 and expanded distribution through Benchling in January 2026, letting biotech users run Lilly trained models for small molecule ADME and tox and antibody developability inside the software where scientists already log experiments. That is a concrete example of pharma turning internal data into external productized AI.
  3 lilly 4 benchling 5 benchling
• Open and semi open research keeps compressing any lead time. AlphaFold 3 showed in May 2024 that one model can handle proteins, nucleic acids, and small molecules together, while later academic work tested where cofolding models still break and new open projects are turning those capabilities into tools more researchers can use. That makes raw model performance harder to defend as a moat by itself.
  2 nature 7 isomorphiclabs 8 nature 9 arxiv

The next durable winners in AI drug discovery will look less like model vendors and more like integrated drug builders. For Genesis, that means turning Pearl into a compounding system, pairing proprietary structural and assay data with partner workflows and internal programs so each prediction improves what gets made, tested, and advanced in the clinic.