AI Just Mapped the DNA of the Dodo Comeback

The Dodo Comeback just got its first treasure map. Colossal, the company working to bring extinct animals back to life, said on October 14 that its new AI platform has scanned the genetic switches behind the dodo's famously oversized beak. The tool, called GenoMorph, flagged roughly 2,300 candidate regulatory elements tied to beak size and shape, giving scientists a starting list of DNA edits to test in the lab, according to Unite.AI.

Colossal Biosciences announced the milestone alongside a wider update on its bird program: a substantially improved dodo genome assembly, an expanded colony of 36 Nicobar pigeons, newly derived primordial germ cells from the dodo's closest living relative, hatched chimeric chickens grown from cultured embryonic stem cells, and a lipid-nanoparticle system for delivering gene-editing tools into bird embryos. Together, the company said in a statement, the pieces form the genomic, cellular, and reproductive foundation needed to turn predicted traits into experimentally validated edits.

How GenoMorph reads extinct DNA

Most gene hunts start with a simple move: compare many genomes from the same species and spot which DNA differences track with a trait. The dodo makes that impossible, because only one dodo genome exists. Colossal built GenoMorph to work around the gap, combining comparative genomics, developmental biology, and high-resolution 3D scans of skulls to predict which regulatory elements were active during development.

The platform was trained on developmental data from chickens, quail, ducks, and pigeons across six stages of embryonic growth, then combined that material with genomes and skull measurements from more than 70 bird species. From that training set, the system prioritized the candidate switches most strongly associated with differences in beak form. The company is beginning functional validation of the candidates and may edit confirmed elements in pigeon cells.

"GenoMorph is designed to work with extremely limited genomic data," Colossal chief science officer Beth Shapiro said in the announcement. For comparison, the company's woolly mammoth program has roughly 100 genomes to draw on, while the dodo effort gets by with a single one. Shapiro also noted that the dodo genome assembly is now about 95 percent filled, with the analysis nearing completion.

The hard parts nobody can skip

A target list is not a dodo. The predictions still need laboratory testing, and even confirmed edits face steep follow-on steps: making many edits inside a bird genome, transmitting those edits to the next generation, and preparing a safe home for the animals on Mauritius, the island where the dodo actually lived. CEO Ben Lamm has described the platform as a treasure map for edits, but also said major hurdles remain before anything hatches.

Skeptics inside and outside genetics will point out that the GenoMorph manuscript has not yet passed peer review, and that lab validation is where AI predictions usually shrink. A regulatory switch can look decisive in a model and turn out to be noise in an embryo. The company's own framing backs this caution: the announced elements are candidates for testing, not proof of which DNA changes shaped the extinct bird's anatomy.

The deeper question is what a revived dodo would actually be for. Supporters argue the project advances conservation tools, from genome assembly methods to embryo editing, that could protect living birds from extinction. Critics counter that the money and attention could go toward species that are vanishing right now rather than one lost in the late 1600s. Both views can be true at once, and the debate tends to grow louder each time Colossal reports a milestone.

Still, the trajectory is hard to dismiss. Last year's advances in pigeon primordial germ cells, gene-edited chicken surrogates, and comparative avian genomes laid the groundwork, and this year's update layers an AI prioritization system on top of a fuller genome. Where the dodo program once had one genome and no map, it now has a near-complete assembly plus a ranked list of switches to test. That is not a bird, but it is a research program with a narrowing to-do list.

For readers, the payoff is years away. Colossal says it aims to produce a dodo in the early 2030s, a target that assumes every remaining step cooperates. What changes now is the bottleneck: instead of hunting blind through sparse ancient DNA, scientists have a prioritized set of edits to run through cells and embryos. Follow the validation work, not the headlines, because the number of candidates that survive testing will say more than the size of the original list.

De-extinction is moving from one-off stunts toward reusable infrastructure. The company built GenoMorph for thin-data species generally, and the dodo is only its first proving ground. If the pipeline works on the extinct bird with the thinnest genetic record, it becomes a template for other lost species. If it stalls in the lab, it joins a long line of AI predictions that looked better on a screen than in a cell.

The story also sits inside a bigger shift in biology, where AI systems trained on morphology and development now guide wet-lab experiments rather than merely describing them. Recent coverage of deep-time science, from deep-earth magma that may have filled our skies with oxygen to the Shanidar spear wound that hints at the oldest known killing, shows how much of modern discovery is data meeting fossils. The dodo is the newest entry in that genre, except the fossil might someday breathe. Meanwhile, the research community keeps testing how environments rewrite biology, as in the microplastics study showing plastic particles are not inert in water.

Watch the next year of lab results. A few confirmed edits in pigeon cells would mark the moment the Dodo Comeback stopped being a database and started being biology. Until then, the treasure map is real, but the treasure has not been dug up.