Karyotype evolution, at the scale of a tree of life.
How many chromosomes a species has, how they're arranged, and how fast that changes varies wildly across life. This part of the lab tries to explain why. We work on it with phylogenetic comparative methods, a long-running CURE program, and a growing layer of AI agents that handle the parts of the job humans shouldn't have to.
What we've built
Dataset Six karyotype databases, 20,000+ species, all machine-readable Coleoptera, Diptera, Amphibia, Mammalia, Drosophila, Polyneoptera.
Each database is curated from the primary literature, checked against voucher citations, and published as both a browsable page and a static JSON export. Other labs and AI agents can consume them without scraping. The Coleoptera database alone has 8,000 species and is the most comprehensive cytogenetic dataset for any insect order.
CUREs program A course-based research program that feeds the dataset 63,682 karyotype records across 55 eukaryotic clades, assembled by an undergraduate cohort in a single semester.
The CUREs cytogenetics dataset covers 63,682 karyotype records across 55 eukaryotic clades, assembled by a single cohort of undergraduate researchers in one semester. Students learn to use AI to locate the best papers and databases with phylogenies or karyotype records, and along the way they encounter how scientific data are actually shared; some tracked down authors directly to request unpublished datasets. The result is a dataset no single graduate student could have compiled, and a model for what undergraduates can accomplish when the course is built around real research problems. This semester, those same 55 clades have been distributed among 50 new undergraduates, each designing and running their own comparative test, starting from a question they found genuinely interesting and working through to a result.
Methods Probabilistic models for chromosome number evolution chromePlus, chromePlusBinary, and related tools for inference on phylogenies.
The models couple chromosome number change (dysploidy, polyploidy) to binary or discrete traits, so you can ask whether karyotype evolution is biased by sex-chromosome state, ploidy, or any other categorical variable. We use them for all of the comparative work on this page and ship them as R packages.
What we've found
April 2026 preprint Dysploidy rates vary 844-fold, and even birds aren't static Copeland, McConnell, Barboza et al., 2026.
Working from the CUREs dataset, we measured rates of chromosome number change across 55 eukaryotic clades. The slowest and fastest clades differ by a factor of 844. Birds, which textbooks treat as the canonical example of chromosomal stasis, sit above the global median once microchromosome dynamics are accounted for. The takeaway: stasis is not a property of a kingdom. It tracks life history.
Drift papers, 2024 For most of animal evolution, drift does more work than selection Two companion papers on beetles and carnivores.
Comparative work across Coleoptera and Carnivora shows that effective population size (or its proxy, range size) predicts karyotype evolution rates better than any measure of selective pressure. The implication is uncomfortable for the classic adaptive accounts of chromosome evolution: a lot of what looks selected is just what happens when populations get small.
Stasis tracks life history, not taxonomic group. Where you sit on the tree is less important than how big your populations are and how long your generations last.
What's next
AI lead investigator An agent carrying a complete project from hypothesis to manuscript We expect it to fail. That's the point.
We gave a Claude-powered agent a real research problem in claw evolution and asked it to run the full arc: form hypotheses, select comparative methods, run analyses, write up. It will fail in places, and each failure teaches us how to design the next version. The goal is not a robot PI. It's an agent that works at our side as scientific collaborator, programmer, and an army of students and postdocs, doing meaningful work when the task is correctly scoped, tested, and examined.
Scaling up Extending the CUREs model to other traits and other clades Morphology, behavior, ecology. Not just chromosomes.
TraitTrawler is the generalization of our cytogenetics agent: a literature-mining pipeline that can be pointed at any trait and any clade and produce a validated dataset. The vision is a future where the bottleneck for comparative biology is not data extraction but research questions. We are actively looking for collaborators who want to run it on their systems.