Define the biological question
Identify a meaningful signal, the constraints that matter, and what that signal leaves unresolved.
The Challenge / Our Approach
Vast biological spaces. Sharper choices.
Biology presents more possibilities than we can examine experimentally. We’re studying whether reinforcement learning can use computational feedback to explore those spaces and focus further evaluation on better-supported candidates.
THE BIOLOGICAL QUESTION
Which possibilities are worth studying?
A useful search starts with a biological question and a clear account of what the available feedback means. Our first focus, peptide–MHC binding, gives us a concrete place to study that question. Computational binding scores can guide exploration, but they do not establish what happens in a cell.
FROM SEARCH TO PRIORITY
Explore broadly. Prioritize wisely.
We’re investigating whether reinforcement learning can use feedback to improve priorities across large candidate spaces. We compare adaptive search with simpler strategies to learn when the added complexity is useful.
Explore candidate space
Use computational feedback to update priorities across many possibilities before choosing what merits further study.
Test whether adaptation helps
Compare adaptive selection with simpler methods and ask whether it produces better-supported priorities.
EVIDENCE AND ITS LIMITS
A promising score is a starting point.
Computational predictions can guide priorities, but they are not measurements of biological outcomes. Fair comparisons must show whether adaptation helps, and biological claims require the right experimental evidence.
See our first research focus