Representation
How can a model learn useful features of biological sequence?
Investigate representations that retain information relevant to a defined biological task, with evaluation beyond the data used to develop the model.
Research / Genomics
From biological information
to better questions.
Exploring how AI can connect patterns in genomic data with hypotheses about biological function.
Genomic data is rich in information. Its meaning depends on context.
Sequence, variation, and molecular activity offer different views of biology. Our research direction explores how models can connect those views without losing sight of the quality, provenance, and limitations of the underlying data.
Questions guiding this frontier
Context
How does a signal change across biological settings?
Explore connections between sequence variation and biological context. Treat an association as a starting point for investigation, not as proof of mechanism.
Prioritization
Which hypotheses are worth testing next?
Study how uncertainty-aware models could help focus experimental attention on questions that are both meaningful and tractable.
The standard of evidence
Keep the claim
close to the evidence.
A model’s output is not a clinical interpretation. Genomic predictions need appropriate validation, and research using human data requires suitable consent, governance, and privacy protections.
The next question
changes everything.
Bring a scientific challenge. Let’s explore what becomes possible when disciplines work together.
Start a collaboration