David Baker, who received the Nobel Prize for protein design, is directing AI systems toward the creation of molecules that do not occur in nature. The move follows his earlier work on computational protein structure and extends it into territory the natural world has not explored.
The interview
Baker addressed the trajectory of biological design in an interview with WIRED en Español. The discussion centered on the next phase of the field and the balance between opportunity and hazard when biology is taken past natural limits. Prior efforts stayed within sequences and folds observed in living organisms; the current direction deliberately steps outside those constraints.
The interview examined where biological design heads next. Baker described the use of AI to generate molecules absent from the natural world and noted both the scale of potential applications and the corresponding risks. No additional technical specifications or performance numbers were released in the conversation.
Expanding the scope of design
The interview makes clear that the tools developed for protein structure prediction and design are now being applied to chemistry that has no precedent in living systems. This is not an incremental extension of existing protein engineering but a deliberate move into synthetic space. Baker framed the change as the logical next step after the field mastered the redesign of natural proteins.
Because the molecules in question have no evolutionary record, standard reference points for stability, toxicity, or interaction with biological systems are missing. The interview therefore paired the discussion of new capabilities with explicit mention of risk. Baker did not present these risks as hypothetical; he treated them as factors that must be addressed as the work proceeds.
Why it matters
The decision to target non-natural molecules with AI changes the starting assumptions for anyone who will later use or regulate the output. Work that once began with sequences drawn from nature now begins with candidates that have no natural counterpart. That shift removes the ability to draw direct comparisons with evolved proteins and forces new methods for safety assessment, environmental impact, and manufacturing control.
For groups already working in synthetic biology or materials science, the interview signals that the technical barrier to producing such molecules has dropped. What was previously limited by the difficulty of designing stable, functional structures can now be approached through AI generation. The result is a narrower gap between what can be imagined and what can be produced in the lab.
At the same time, the absence of natural precedents means that existing regulatory and review frameworks, which often rely on analogy to known biological molecules, will have fewer anchors. The interview does not resolve how those frameworks should adapt, but it indicates that the question is no longer theoretical. Teams responsible for biosafety, intellectual property, or downstream application will need to update their models sooner rather than later.
The concrete outcome is that design work is moving from mapping and modifying what evolution has already produced to generating structures that evolution never tested. The interview records that transition as already under way.
---
Sources:
{"word_count": 612}
No comments yet