In population genomics, there are a dizzying array of potential data analysis approaches to infer population history, aspects of natural selection, or other evolutionary properties from data. Although methods developers try to evaluate their approaches, those evaluations can be unconsciously biased or may not reflect the experiences of real-world users. GHIST is an annual forum in which the community can test inference approaches in an unbiased fashion. Each year, the GHIST organizers release simulated population genomic data sets and host a competition to infer various aspects of the processes that generated those data. From the competitors’ solutions, the community will learn what approaches perform well or poorly in particular circumstances. Participating the competition is also great training for new students and researchers!
GHIST1 has now concluded. It consisted of four demographic history inference challenges of escalating expected difficulty. For each challenge, we provided a VCF file with genomic data, and competitors submitted a simple text file with their inferences. We had roughly 60 registered participants, with 47 submissions to the simplest challenge and 12 to the most complex. We'll be reporting the full results at ProbGen, and writing a manuscript in early 2025. The top will be co-authors on the publication describing the competition.
GHIST2 will be launching in summer 2025. We will expand the challenges beyond demographic history inference, including at least sweep detection. We are actively seeking community thoughts about how to structure the 2025 challenge.
To see the first competition, visit its page on Synapse. (You will need to create a Synapse account.)
To help competitors get started, we created an hour-long introductory workshop. In the workshop, we introduced GHIST, used dadi-cli on a cloud instance to analyze the data from the first challenge, and submited our inference to the tournament. Join in!
Questions? Contact Ryan Gutenkunst at rgutenk@arizona.edu.
Thank you to the Design Committee:
- Katie Lotterhos - Northeastern University
- Andrés Moreno-Estrada - Centro de Investigación y de Estudios Avanzados del IPN
- Peter Ralph - University of Oregon
- Adam Siepel - Cold Spring Harbor Laboratory
Also thank you to Travis Struck, who implemented the 2024 competition.
This competition is supported by NIH NIGMS grant GM149235 to Ryan Gutenkunst.