Learning thelogic of life.
We build AI that reads, simulates, and reasons about biology, from single cells to the human brain, and we test what it predicts in the lab.
A language model proposed a cancer therapy. The experiment agreed.
We taught a language model to simulate biology. Built with Google DeepMind, our 27B-parameter model ran thousands of virtual drug experiments and flagged one that could make tumors visible to the immune system. Experiments in human cells confirmed it.
Five questions
we keep asking
We combine mathematics, machine learning, and large-scale biomedical data. Our models are judged by one test: whether what they predict turns out to be true.
- 01 Biology as a language We turn cells, tissues, and genomes into text so that language models can read, write, and reason about them. →
- 02 Virtual cells & causal perturbation Predicting what a drug or a gene knockout will do to a cell before anyone runs the experiment. →
- 03 Foundation models of the brain Models trained on thousands of hours of brain activity that predict clinical traits and simulate neural dynamics. →
- 04 AI agents for science Teams of language-model agents that plan analyses, pull data, generate hypotheses, and design experiments. →
- 05 Foundations of generative models & intelligence New generative models for sequences, sets, and continuous dynamics, and a theory of where reasoning comes from. →
Lately
MoRSE accepted at NeurIPS 2026
Peiwen Li, Harry Zhang, Yangtian Zhang, and Sizhuang He introduce a mixture of role and subtask experts that lets teams of LLM agents split up hard problems.
The AI Neuroscientist
Aakash Patel and colleagues release an agent that runs neuroimaging analyses through conversation, with the Saxena and Krishnaswamy labs.
Teaching language models to simulate biology, not just reason about it
David speaks at the EMBL-EBI Industry Programme workshop on AI reasoning with LLMs, Cambridge, MA.
STRIDE presented at ICML 2026
Syed Rizvi presents STRIDE, which post-trains LLMs to reason about and refine biological sequences through edit trajectories.
Aakash Patel named a Salovey–Moret Data Science Fellow
Aakash joins the inaugural cohort of Yale’s Peter Salovey and Marta Moret Data Science Fellows Program.
Recent papers
Around 80 papers since 2010, in venues from Cell and Nature Methods to NeurIPS, ICML, and ICLR. Browse everything →
- 2026 NeurIPS 2026
- 2025 bioRxiv Code / weights ↗
- 2025 NeurIPS 2025
- 2025 ICLR 2025 Code / weights ↗
Models & code you can use
We release weights, code, and data. Our C2S-Scale models are downloaded thousands of times a month.
- Google Research
- Google DeepMind
- Wu Tsai Institute
- Boehringer Ingelheim
- NIH / NIGMS
- National Science Foundation