What is Evo 2 and what can it do?
A new class of genome‑scale AI
Evo 2 is a biological foundation model trained on an unprecedented volume of DNA sequence data. Built by combining deep learning with massive genomic datasets spanning many species, the system learns statistical patterns of genomes and can perform a range of sequence‑level tasks.
What the model can do
- Identify genomic elements such as genes, splice sites, and other regulatory signals.
- Predict functional annotations from raw DNA sequence.
- Generate short sequences that fit learned biological constraints, opening a path toward in silico design of genetic parts.
Potential applications
- Research acceleration: automating routine annotation and hypothesis generation across bacteria, plants and animals.
- Synthetic biology: proposing candidate sequences for gene circuits, promoters or metabolic pathways that researchers can test in the lab.
- Diagnostics and discovery: highlighting unusual sequences or regulatory motifs linked to disease or environmental adaptation.
Opportunities and limits
While the model marks a major advance in scale and capability, it does not yet replace experimental validation. Predictive power depends on the diversity and quality of training data, and generating whole, viable genomes remains beyond current, safe practice. Ethical and safety considerations are central: models that suggest functional sequences must be governed by access controls, oversight, and clear lab‑testing protocols.
In short, Evo 2 brings AI tools to genome‑scale problems, boosting the pace of discovery and design while shifting the emphasis onto responsible deployment and rigorous experimental follow‑up.