Is Evo 2 the Future of Biology?

Developed recently by scientists in the Arc Institute and collaborators from Stanford, UC Berkeley, and UC San Francisco, Evo 2 is an AI model purposed to make predictions in biological fields––and it may be the most promising one yet.

Evo 2 was trained on DNA data from every living species on earth––from prokaryotes to plants and animals––unlike its predecessor, Evo 1, which was only fed single-cell genome data.

Well, why is it so special? Evo 2 functions similarly to other LLMs (Large Language Models), in that it uses a series of algorithms to predict the next word––or, in this case, the next nucleotide in a DNA sequence. In this manner, predicting gene sequencing, which could take researchers years to uncover, can be done with a simple prompt. These findings can then be used to make malignant/benign classifications for mutations, allowing scientists to model an accelerated evolutionary process, where mutations either harm or benefit, or do not affect an individual.

What excites me most about the model, however, is the potential with which future research projects can build off of it. Evo 2 is an excellent standalone tool for making predictions or classifications, but what's more important is how this information can be applied to treatments. I believe that, currently, its best application would be to aid CRISPR treatment.

CRISPR is a gene-editing treatment that seeks to target and alter (or cut out) harmful genes. Though it has seen great success with common, shared diseases like sickle cell, it is a rarer solution for individual-specific issues, which require treatment personalization, and thus, higher risk. As stated by the Children's Hospital of Philadelphia (CHOP), "relatively few diseases benefit from a 'one-size-fits-all' gene editing approach." However, just days ago on May 15, 2025, a baby at CHOP became the first successfully treated individual for personalized CRISPR, a healthcare milestone. It's risky, but we now know it's possible.

How does Evo 2 tie into all this? Well, the AI model is exceptionally good at identifying which mutations in the genome are harmful, making it much easier to pinpoint where things went wrong. CRISPR relies on specificity to ensure that RNA correctly tracks the right target sequence to edit.

If, given a patient's data sequence, Evo 2 can predict exactly where errors occur, then perhaps CRISPR can be applied to a specific disease in a much safer manner.

Evo 2 and specialized CRISPR treatment are quite new to the table, but I am excited to see where their intersection can lead us. Even more, I am glad to see collaborations between data science and biology making great advancements, and am sure it will continue to do so if we embrace the syntheses of these fields.

Credits:

Children’s Hospital of Philadelphia. “World’s First Patient Treated With Personalized CRISPR Gene Editing Therapy at Children’s Hospital of Philadelphia.” Chop.Edu, 15 May 2025, https://www.chop.edu/news/worlds-first-patient-treated-personalized-crispr-gene-editing-therapy-childrens-hospital.

Myers, Andrew. “Generative AI Tool Marks a Milestone in Biology.” Stanford.Edu, Stanford University, 2025, https://news.stanford.edu/stories/2025/02/generative-ai-tool-marks-a-milestone-in-biology-and-accelerates-the-future-of-life-sciences.