Colossal's GenoMorph Aims to Solve De-Extinction's Missing-DNA Problem
Colossal Biosciences announced GenoMorph on Oct. 14, 2026, an AI platform to speed up de-extinction when ancient DNA is scarce, starting with the dodo bird.
GenoMorph is a target-discovery system for genome engineering, not an editing tool itself. Chief executive and co-founder Ben Lamm said the build took more than a year of work. TechCrunch
Why the dodo is a hard starting point
Colossal is working with only one dodo genome, compared with 100 genomes for its woolly mammoth project. The company attributed the gap to preservation. Woolly mammoth DNA is more plentiful because cold climates preserve DNA better than the tropics where the dodo lived.
How GenoMorph works
GenoMorph combines 3D morphology, the precise measurement of physical shape, with comparative genomics and developmental biology to find genes to edit in species with limited gene data. In practical terms, it fuses shape data, DNA comparisons across species, and embryo-development data matched by growth stage to rank candidate DNA regions for lab testing.
To build it, Colossal collected developmental data from chickens, quail, ducks, and pigeons and combined it with genetic information and 3D skull measurements from more than 70 bird species. The logic is comparative. When the target species offers only a single, possibly incomplete ancient genome, living relatives provide growth patterns and links between genes and traits that researchers can apply with caution.
The first reported output was about 2,300 promising DNA editing targets for dodo cranial facial morphology, for lab testing. The focus is head and face shape. Skull shape can be measured in 3D and compared across birds, which gives the model a clear physical trait to learn from.
From computer predictions to live birds
Colossal also said it developed a system that uses lipid nanoparticles, tiny fat-based carriers, to deliver gene-editing tools directly into developing bird embryos. Delivery inside the egg is difficult because the injection methods used in mammals do not transfer to birds.
Parallel work on the surrogate continues. The company said its Nicobar pigeon colony now has 36 birds and that it created the essential cells that become eggs and sperm in Nicobar pigeons. The Nicobar pigeon is the dodo's closest living relative, which makes it the host species for dodo edits.
The business background
Colossal calls itself "The De-extinction Company" and says it will be the first company to use CRISPR technology successfully for de-extinction of lost species. It describes its de-extinction program as a $615M program. The company opened its Dallas genome engineering lab to media for the first time in March 2026. Lamm discussed AI and de-extinction with CNBC's Julia Boorstin at the 2026 CNBC AI Forum on Oct. 2.
Colossal was reportedly in talks in July 2026 to raise new capital at a $20 billion to $30 billion valuation, which would have doubled or tripled its previous valuation. TechCrunch
For scale, Romulus, Remus, and Khaleesi were produced through 20 gene edits in gray wolf DNA and born via surrogate dogs at a 2,000-acre preserve. Four red wolves were cloned using recovered ghost alleles.
The broader context here will be familiar to readers who follow biological AI models. Sequence data alone does not determine physical shape, especially with a sample of one. Colossal is betting that structured shape data and development timelines can narrow the search enough to make editing of many genes workable.
In my view, it is worth flagging what GenoMorph does not yet solve. Ranking 2,300 candidates is progress. Testing them in bird embryos, passing changes to offspring, and producing viable surrogates is still slow and costly biology. The nanoparticle delivery system and the Nicobar pigeon egg-and-sperm work matter as much as the model, because predictions without a fast edit-and-test cycle stall.
Over the longer term, there is reason for optimism. De-extinction has long been limited by how preservation varies by climate and species. If a reusable platform for low-data species extends beyond dodo skull shape to other traits and birds, it would lower the cost of each later attempt. That is the outcome to watch, more than any single target count.
