Prima MenteAlzheimer’s Disease Data Initiative (ADDI)
Registration openStarts spring 2027

A global AI competition to find new therapeutic hypotheses for Alzheimer’s disease. Build models upon a new open source cellular atlas of 150M neurons, astrocytes, and microglia. The best hypotheses will be validated in our wet lab.

Register now to get updates on the competition details, tasks and prizes, and access to the atlas when released.

By registering, you agree to our privacy policy.

In collaboration with
  • NVIDIA
  • Hugging Face
  • Nebius
  • Talisman
  • Ultima Genomics
  • Cellanome
  • Prime Intellect
  • Goodfire
  • Boltz

The Challenge

Build AI models on a new cell atlas, to develop novel hypotheses that we will validate in our lab.

Cells
150 million cellsneurons, astrocytes and microgliaspanning healthy and disease contexts
Perturbations
Combinatorial chemical and genetic perturbationswith multimodal readouts
Competition starts
Spring 2027Atlas released to all participants

Two stages

Stage One. Each Participant will build AI models upon the atlas, and will be assessed against a set of evals: general benchmarks, harder tasks such as phenocopy, and a selection of evals proposed by the community. Participants with the strongest results are shortlisted for Stage Two.

Stage Two. Utilising their models, each shortlisted Participant will propose one new therapeutic hypothesis for Alzheimer’s, together with the molecular and functional phenotypes anticipated. Finalists’ hypotheses are tested in our wet lab, closing the loop from biology to target to function.

Timeline

We’ll share more milestones as we go.

  1. October 2026Registration opens

    Atlas generation is underway.

  2. November 2026Community launches

    A place to ask questions, follow updates, and find teammates.

  3. December 2026Prizes announced

    We will share the prizes for both stages.

  4. Spring 2027Competition starts

    The dataset is released. Build AI models and get assessed on a curated set of benchmarks and complex evals.

  5. Summer 2027Hypothesis validation

    Shortlisted participants propose a therapeutic hypothesis. Finalists will have theirs tested in our lab.

Q&A

Anyone! You could be coming from AI, biology, the clinic, or beyond. You can enter on your own or as a team. We will launch a community soon to help you find teammates. Academic and industry participation is encouraged!

No. The task is set up so machine learning researchers can take part without wet-lab experience. However, biology experience will certainly be advantageous. We will provide learning resources as well as a community to ask questions and share insights.

Yes. The dataset will be open source, and you will use it to build your models in the first stage. We will share details on access before the challenge starts. The data will be made available through AD Workbench and Prima Mente’s modelling platform.

We are generating the cells now. The challenge starts in spring 2027, once the full atlas is ready. We will release data in batches, so that we and others can also learn along the way.

Yes. When you register, use the optional question to tell us which gene or compound you’d like to see tested in the lab, and why. We will review every suggestion.

Yes! Prize details will be announced in the coming weeks.

Models will first be assessed on a curated set of evals, across general benchmarks, more complex tasks, and a selection taken from community proposals. We understand that evals can oftentimes be contrived or not representative of real biology or good modelling. We wish to take this very seriously, and will allow for public comment on the set of evals used for assessment.

The strongest Participants will then be shortlisted for the second stage and asked to propose a single therapeutic hypothesis with anticipated molecular and functional phenotypes. Finalists will have their hypotheses tested in our lab, and winners selected across a set of outcome criteria. More on judging and prizes will follow in the coming weeks.

No, the challenge is free to enter.

No. Prima Mente does not obtain any rights to any model or discovery you make by participating in the Challenge.

We will launch a community closer to the time for questions about the challenge, regular updates, and finding teammates. Everyone who registers will hear when it opens. For anything else, email challenges@primamente.com.

In 1906, Alois Alzheimer described a new disease of the brain. After 120 years of progress, we still don’t understand what drives Alzheimer’s.
We want to change that.

Register