Colossal Narrows Dodo Revival to 2,300 DNA Targets With New AI Tool

Colossal Biosciences says it has identified 2,300 DNA targets tied to the Dodo's defining traits, using a new AI tool called GenoMorph that the company announced on Wednesday.

The number matters because Colossal has exactly one Dodo genome to work with. The company, founded in 2021, claims to have already resurrected the dire wolf. But the Dodo presents a harder problem: its process must run in an egg, not in a surrogate gestating a mammal. The flightless bird, endemic to Mauritius and documented by the Portuguese in the 16th century, disappeared after invasive species and human hunting wiped it out.

Colossal also said it used lipid nanoparticles to conduct genomic editing inside an egg for the first time.

What does GenoMorph actually do?

Chief Science Officer Beth Shapiro framed the core problem in concrete terms: how do you edit a Nicobar pigeon — "a smaller, beautiful-colored, strongly flying pigeon with a tiny little beak" — into the iconic dodo when you have a single reference genome?

For the mammoth, Colossal can compare roughly 100 mammoth genomes against elephants and spot where they diverge. For the Dodo, that comparison is impossible. GenoMorph fills the gap. Shapiro's team combined:

• DNA sequence data from a range of bird species
• 3D skull scans summarized by machine learning
• Developmental data on how bird faces form at early stages
• Gene expression data showing which genes switch on and off, and when

Fed all of that, GenoMorph outputs a candidate list of edits. "Every cell has the same book in it, but which chapters should I be reading if I want to make a dodo beak versus a Nicobar pigeon beak?" Shapiro said.

The takeaway: Colossal has turned a single genome into a ranked, testable edit list.

Is this really a Dodo?

No — not by genetic replication. Colossal is building what its leadership calls a 'functional' Dodo: the genome of a living pigeon relative, edited with Dodo DNA sequences tied to specific traits like the bird's beak. Some scientists have strongly objected to the company's framing.

Shapiro's answer is that species boundaries are looser than critics admit. "It's a species definition that's based on an arbitrary threshold of sequence similarity," she said. "But that threshold is arbitrary. It's going to be arbitrary no matter who you are."

CEO Ben Lamm put it more bluntly, citing critics online who argue a duck that looks, quacks and eats like a duck still isn't a duck: "Tell my two and a half year old that."

Lamm also noted the numbers behind the mammoth project: 104 mammoth genomes spanning 1.3 million years of genetic divergence, none yielding living cells. "You know what's crazy about 1.3 million years of genetic divergence? They're very different. You know what's not different? The parts."

How does the AI fit with frontier models?

GenoMorph is not an LLM. Shapiro said frontier models "are not really trained to solve this particular problem," which is mapping a genotype to a phenotype — connecting developmental biology to how an organism ends up looking.

Lamm said early LLMs couldn't handle ancestral state reconstruction or comparative genomics, so Colossal built its own. The company has access to "the latest and greatest" from OpenAI and Anthropic, and uses it for certain tasks — but not for GenoMorph yet. He said the thinned-down data sets and methods could later feed into superintelligence platforms.

The takeaway: Colossal's proprietary data, not a frontier model, is the asset here.

What counts as success?

Shapiro defined the goal in collaboration with the Mauritian Dodo Advisory Committee: a large, ground-dwelling, fruit-eating bird that restores the Dodo's ecological niche. "If we are successful in doing that, we will have brought back the dodo," she said.

She pointed to the dire wolves as precedent: they're white because ancient dire wolves were white, but Colossal used a white-coat variant circulating in today's gray wolves, after evidence suggested the dire wolf version could have left them deaf or blind in a gray wolf genetic background.

Lamm argued the ecological bar is the harder one to clear, not the cosmetic one. "If I just had to make a bird that looked like a dodo, my life would be easier, and I could raise way less money."

Colossal has also announced quests to bring back the mammoth, the thylacine and the giant Moa — and Lamm said the core GenoMorph platform may move faster across those projects than expected, something he said he learned only that morning.