Reading lab records to suggest next experiments
Could the data tell us which experiment to run next?
Even with notebooks full of results, we think the next experiment still depends largely on a researcher's intuition. We plan to structure those records and show candidate next experiments with the reasons behind them. As with KITE, which we have published, we would like this work to end up as a tool anyone can pick up.
Lab notes and result data
We aim to connect and clean notes and results, then have agents propose hypotheses and verify each other.
A proposed next-experiment design, with reasons
| Field | Biology |
|---|---|
| Status | Building pipeline |
| Input | Lab notes · result data |
| Output | Next-experiment design |
| Delivery | Joint research |