OpenScientist AI Completes Five Published Research Projects

C-BRAIN's Nature Medicine study details how the platform performed on real scientific analysis, including Alzheimer's-related trial data.

Updated on Nov. 21, 2026 in Alzheimer’s

OpenScientist AI Completes Five Published Research Projects

C-BRAIN published a study in Nature Medicine on November 16 describing five scientific projects completed by OpenScientist, an AI platform developed by scientists at Lawrence Berkeley National Laboratory. The study's lead authors were Justin Reese of Lawrence Berkeley National Laboratory and Randall Bateman of Washington University in St. Louis.

Why it matters

The study shows an AI system and human analysts reaching the same primary conclusion on Phase 3 trial proteomic data, while the platform's open design lets users track and validate its reasoning and code. For readers following AI-assisted research, it offers a concrete, checkable example of how such tools are being evaluated.

The peer-reviewed paper reports five completed projects, building on four examples from an April 2026 preprint, with three more case studies in the supplement. In the Phase 3 suvorexant comparison, both the AI and human analysts found no protein changed significantly in response to treatment.

The players

C-BRAIN

C-BRAIN is the organization that developed OpenScientist with scientists at Lawrence Berkeley National Laboratory and is offering financial support for five discovery partners.

Lawrence Berkeley National Laboratory

Lawrence Berkeley National Laboratory is a U.S. Department of Energy research facility in California where the OpenScientist platform was developed.

Randall Bateman

Randall Bateman is a researcher at Washington University in St. Louis who co-led the Nature Medicine study on OpenScientist.

The details

OpenScientist conducts iterative research rounds, including literature searches, hypothesis generation and user feedback, before producing a final report. Human analysts grouped proteins into biological pathways, while the AI identified statistically significant relationships between protein groups across time points.

Timeline

  1. April 2026: A preprint covering four OpenScientist examples was covered by Alzforum.

  2. November 16, 2026: Nature Medicine published the OpenScientist study.

  3. December 15, 2026: Deadline to apply for C-BRAIN discovery partner support.

Big Picture

The study places AI-assisted analysis alongside human analysts on the same Phase 3 suvorexant proteomic dataset, testing whether automated reasoning can reproduce the human primary conclusion. It follows the pattern of AI tools being checked against established human-led scientific workflows rather than accepted on claims alone.

For people following Alzheimer's research, the study points to a new kind of analytical tool that may speed how trial data is examined, though the platform is not a treatment. Readers interested in the platform itself can try OpenScientist through C-BRAIN's site.

The takeaway

Tools that analyze research data are most useful when their reasoning can be checked, and this study shows one approach to making that possible. Readers can watch how such systems perform against established human analysis as more results are published.

What happens next

C-BRAIN is accepting applications for five discovery partner positions, with applications due December 15.

Further reading

For more on the disease at the center of several of these analyses, see our Alzheimer’s section.

More information

You can try the platform yourself on the C-BRAIN site.

Source note: This article includes information reported by Alzheimer Research Forum.