David Liu Team Redesigns Botox Enzyme With AI, Boosting ALS-Linked Protein Cutting 80-Fold
Updated
Updated · Singularity Hub · Jul 28
David Liu Team Redesigns Botox Enzyme With AI, Boosting ALS-Linked Protein Cutting 80-Fold
1 articles · Updated · Singularity Hub · Jul 28
Summary
ProteinMPNN-generated Botox-enzyme variants gave David Liu’s team a stronger starting point for directed evolution, producing enzymes that cut an ALS-linked protein nearly 80 times more efficiently and 56 times more selectively than versions evolved from natural enzymes.
58 AI designs were screened and the top three proved highly soluble and sometimes more active in E. coli, addressing a core problem in enzyme engineering: natural starting proteins often destabilize and fail as mutations accumulate.
The team then ran the redesigned enzymes through PACE—Liu’s rapid evolution system—to retarget botulinum neurotoxin proteases toward a mutant region of a neuron-related protein that can clump and damage cells in ALS.
Tests were limited to immortalized human cells, so performance in more complex biological settings is still unproven, but the work suggests AI-stabilized proteins could widen the range of designer enzymes beyond what natural templates allow.