AI as a research copilot.

IRED-88 is an imine reductase from an enzyme engineering campaign Eric worked on at Novartis, published in ACS Catalysis in 2021. The campaign measured 4,720 single mutants, the enantioselectivity of 310 engineered variants, and a crystal structure. Next to it sits a six-paper knowledge base in the format you built in 3.2.

1. Which mutations improve activity? The deep mutational scan shows which single mutations help, and where they sit along the sequence.
2. Are the hits where the structure says they should be? Each position’s distance to the NADP cofactor in the crystal structure.
3. What does prior work already know? The top hits, checked against the knowledge base.
4. What does prediction see that the crystal cannot? ESMFold fills in the residues the crystal leaves out, with a confidence score attached.
5. Additive or epistatic? Doubles and triples of the best mutations test whether a claim from the literature survives the data.
6. Would we have noticed a hole? Hide part of the scan where the literature matters, and see what the rest of the evidence recovers.

Read every notebook, with its code and results.

The demo page shows how to download the notebooks and their data and run them on your own computer. You can also hand the folder to the agent you set up in 3.0 and ask it questions about the data, the same way you questioned your knowledge base in 3.2.

Two minutes of feedback, so the next session is pitched better. It also records your attendance, so fill it in whether you were in the room, on Zoom, or catching up on the recording.

Session 3 feedback →