The next chapter of our Obesity ML work with the Eric and Wendy Schmidt Center is live. After mapping single-gene perturbations in Crunch 1 and double-gene perturbations in Crunch 2, we turn to the harder question. Which combinations of genes, drawn from millions of possibilities, will most strongly drive thermogenic activity in adipocyte cells?

The problem

Obesity remains among the most consequential health challenges of our time. Thermogenesis, the process by which cells convert energy into heat, is a promising therapeutic axis. The Eric and Wendy Schmidt Center at the Broad Institute is asking the Crunch community to help nominate the gene pairs most likely to activate this process when perturbed together.

Four Million Possibilities

Participants will evaluate all possible pairs drawn from 2,991 genes (those above 5 transcripts per million, plus transcription factors and marker genes). This yields 4,474,413 candidate combinations after filtering out pairs already tested in Crunch 2. For each pair, predict a score for each thermogenic gene signatures, and then rank them.

The Wet Lab Experiment is the Leaderboard.

This competition does not score predictions in real time. The final ranking will only be determined after laboratory experiments are performed at the Broad Institute. 150 unique gene pairs will be selected for experimental validation, split across two routes:

• ~110 pairs nominated by top-performing Crunch 2 teams
• ~40 pairs nominated by Crunch 3 peer-review winners
• Both routes require participants to peer review each other’s work to keep idea quality as the primary criterion for selection.

The Timeline

• Each participant can participate in both routes and submit two proposals.

• Crunch 2 participants are especially encouraged to adapt their existing models and submit within the first 3 weeks of the competition. 
• After that, all participants, including both Crunch 2 and new participants, have an additional 3 weeks to submit or revise their work before the competition closes.

The Prize Structure

• Evaluation 1 (peer review): 10% of the prize pool. Winners announced at the end of Summer 2026.
• Evaluation 2 (wet lab experiment): 90% of the prize pool. Awarded to the top 10 teams whose gene pairs achieve experimentally validated scores exceeding the maximum perturbation score observed in Crunches 1 and 2. Winners announced in Fall 2026.

Peer Review

To qualify for prizes, you must review 3 to 4 submissions from other participants. This is how the community selects which gene panels make it to the bench. Strong rationale, novel design, and clean formatting are what reviewers are asked to look for.

The data is ready. The lab is waiting: https://hub.crunchdao.com/competitions/broad-obesity-3