We like to call our lab "interdisciplinary and non-traditional", as we approach the science with a core question and mission of understanding and preserving nature, and we use different disciplines to address it. This also means that we need to collaborate, so we often conduct large projects together, Below some of our latest lines of work depicted at different scales:
Genetic diversity is the raw material of adaptation, yet it is nearly invisible to conventional biodiversity monitoring. We build the tools to see it — methods that quantify how much evolutionary potential disappears as habitats shrink, deep learning that maps thousands of plant species from aerial imagery, and century-long genome time series read from herbarium collections. Next: genomic observatories, wild populations resequenced year after year, so adaptation in nature is measured directly rather than inferred.
Key publications: Expósito-Alonso et al. (Science 2022) · Gillespie, Ruffley & Expósito-Alonso (PNAS 2024) · Lang et al. (Nature Ecology & Evolution 2024) · Mualim et al. (PNAS 2026) · Lin et al. (marApp, bioRxiv 2025) · Selmoni et al. (Trends in Genetics 2024) · Huynh, Gillespie et al. (ECCV 2024)
There is only one Earth and no control — so we build the replay ourselves. Through GrENE-net, a crowd-sourced consortium, the same genetically diverse Arabidopsis population was sown at more than 30 field sites in 17 countries and sequenced every generation for five years. Populations adapted within a handful of generations, how predictably they evolved depended on local climate, and those heading for extinction evolved erratically beforehand. A dedicated field evolution station at UC Berkeley is next.
Key publications: Wu, Bellagio et al. (Science 2026) · Expósito-Alonso et al. (Nature 2019) · Leventhal et al. (bioRxiv 2025) · Leventhal, Ruffley & Expósito-Alonso (Annual Review of Plant Biology 2025) · Czech et al. (grenepipe, grenedalf, Bioinformatics 2022/2024) · More at grene-net.org
A population cannot evolve toward a solution whose genetic ingredients never arose. So we write the missing variation ourselves: deep learning models trained on natural variation nominate mutations expected to help under future climates, CRISPR creates them, and fitness experiments judge every prediction. Heritable DNA barcodes are next — letting whole populations of engineered variants compete and evolve over generations, so rapid adaptation can be re-created from parts we chose.
Key publications: Ruffley et al. (bioRxiv 2024), Kyung et al. (bioRxiv 2026)
Join us in advancing our understanding evolution of nature. Click the button below to contribute to our ongoing efforts.