We recently sat down with the five techbio companies accepted into the Fall 2026 Drive cohort to learn more about their science, their “ah-ha” moments, and how Drive is changing the trajectory of their business. Next up, it’s Neoclease, a techbio company using generative AI to design compact gene editors, one target at a time. We spoke with CEO and Cofounder, Chelsea Trengrove.
1. What problem are you obsessed with solving—and why does it matter right now?
For twenty years the field has taken one or two bacterial gene editing proteins and asked every target to accommodate them, which I’ve always thought was backwards. A gene edit is permanent, and an off-target cut isn’t something you can wash out, so we design a new editor for each gene and build the precision in at that specific site rather than trying to engineer it back in later. Our editors are also less than half the size of the standard one, so delivery stops being a hurdle that can kill in vivo programs. We’re starting with LRRK2 in Parkinson’s, and that’s the promise I’d love to see gene editing finally deliver on: one dose, permanent, and genuinely disease modifying.
2. What’s a recent moment that made you think, “this could really work”?
Going from 32 percent to 45 percent editing efficiency in a single design cycle. Nothing about the chemistry changed; we just fed the wet lab results back into our model and the next batch of editors came out better. The in silico design loop was working! That was the moment Neoclease stopped feeling like a set of proteins we had made and started feeling like an engine that makes them.
3. Since joining Drive, what’s one shift in your thinking, strategy, or network that’s had a real impact on your company?
The biggest shift was moving our value inflection point to development candidate. With five mentors in the room together every week, we went through the preclinical plan study by study, asking what each one would actually tell a partner or a regulator. That pushed us to get much sharper about where we’re genuinely differentiated, and to build the package around proving exactly that.
