FROM A POSSIBILITY TO A TESTABLE QUESTION
The research studio.
A proposed program for focused biology research using AI. Every project should have a clear question, a comparison, and a result that can change our minds.
Project 001 / Enzyme performance
Proposed pilot
THE QUESTION
Can AI help us choose better candidates?
For a defined industrial enzyme task, does an AI-assisted selection method identify better-performing candidates than a conventional method using the same testing budget?
What would count as progress?
- A predeclared functional performance measure.
- A fair baseline and held-out evaluation.
- Laboratory confirmation, with uncertainty reported.
- Transparent cost per useful candidate.
THE PATH TO EVIDENCE
Define the question
Choose one industrial enzyme task with a prospective laboratory partner. Specify the conventional baseline, assay, and success criteria before selecting candidates.
ProposedSecure the foundation
Agree the laboratory scope, budget, data rights, intellectual property, and publication terms. Obtain any required approvals and fund the work.
Not startedCompare predictions
Evaluate the AI-assisted approach against the baseline using a held-out dataset. Record uncertainty, compute cost, and unsuccessful candidates.
Not startedTest and report
Have qualified researchers measure selected candidates. Report whether the approach improved outcomes, including negative results and study limitations.
Not startedBuild the program with us.
We are interested in conversations with laboratories, researchers, and company investors. A partner, budget, timeline, and commercial rights have not yet been agreed. No experimental or clinical results are claimed.