A few years ago, we were just a group of researchers chasing unconventional bets in AI.
That journey took us from new architectures like Hyena, to Evo and creating the field of generative genomics, to a much bigger realization.
We need AI that can understand biology as a whole. We started @RadicalNumerics to pursue that vision.
Here’s a conversation with my cofounders on how we got here, why we took the leap, and what we believe comes next.
1/ With Omnii Diagnostics, we asked whether a multimodal genome language model could help find cancers whose signals in blood are too faint for current classifiers.
We adapted Omnii-1 to a biomarker never seen in pretraining. In a test cohort, Omnii's features improved early-stage cancer detection
as we have seen with the big LLMs, with scale + architecture + training generalist models outperform specialized models *even on the task these models specialize on". True for coding...math...and increasingly true in bio 🧬
1/10 Introducing Omnii Genome, one of eight Advances built on Omnii-1.
A single generalist model now outperforms leading specialized models (AlphaGenome) in linking genetic changes to gene activity and mapping regulatory interactions.
These capabilities help make sense of the 98% of the human genome that doesn’t code for proteins
How Omnii-1 works:
1/10 Introducing Omnii Genome, one of eight Advances built on Omnii-1.
A single generalist model now outperforms leading specialized models (AlphaGenome) in linking genetic changes to gene activity and mapping regulatory interactions.
These capabilities help make sense of the 98% of the human genome that doesn’t code for proteins
How Omnii-1 works:
@Noahpinion speaking from the "inside": very important to keep defensive capabilities at the frontier. in this case, we are trying to defend against antibiotic resistant superbugs - unfortunately a real and worsening threat
For sixty years, we could only read the code of life. Now @RadicalNumerics has used AI to design a phage for a patient’s drug-resistant infection in weeks. Congratulations, @exnx and team on the launch of Omnii-1.
At a time when many seek real impact from AI, I'm excited by our team's progress! Moving beyond in-silico benchmarks to practical biology and clinical applications is a major leap. With Omnii-1, we're already seeing strong performance w/ minimal tuning. Excited for what's next!
We built Omnii-1 from the ground up with generative multimodal capabilities. Omnii-1 designed phages are being used to treat antibiotic resistant bacterial infections today, and this is just the beginning.
More at https://t.co/dS7AHbCrE3 - diagnostics, genetics, rare earths!
Thrilled to share the news after a monumental team push! After pretraining Omnii-1 this summer, we're seeing strong performance across numerous application areas, with surprisingly minimal tuning. We're just getting started and I'm so excited to keep building with this team 🧬
Omnii-1 is out.
We've learned a lot about what happens when you train a single model across biology rather than solving each problem separately.
Some of the more surprising results haven't made it into a blog post yet.
Incredibly proud of the team that built this!
Based on benchmark performance for regulatory genomics, E2G linking & variant effect prediction, this model looks SOTA. Not enough details to know exactly what its doing. Seems conceptually similar to Nona https://t.co/Hs3ZVotJpQ. Look forward to testing it when it is available.
Introducing Omnii-1, our most powerful genome language model, and a major step toward general biological intelligence
In a world first, Omnii-1 designed the first AI-generated bacteriophages to be administered to a human patient with an antibiotic-resistant infection under compassionate use
Introducing Omnii-1, our most powerful genome language model, and a major step toward general biological intelligence
In a world first, Omnii-1 designed the first AI-generated bacteriophages to be administered to a human patient with an antibiotic-resistant infection under compassionate use