Direct discovery · Biology · Chemistry · Medicine
Neural iterative design creates high-affinity drug-binding proteins
De novo proteins that bind exatecan and apixaban, with experimentally measured nanomolar-to-picomolar affinity and an exatecan binder that protects the drug from hydrolysis.
Summary
A neural iterative selection–expansion workflow coupled learned protein-sequence design with protein–ligand structure prediction to create de novo binders for exatecan and apixaban. Experimental results reported high success rates and affinities, including a modified exatecan binder that stabilized the drug against hydrolysis.
AI role
Two neural networks iteratively designed protein sequences and predicted protein–ligand co-structures to optimize small-molecule binding.
Narrative role
This expands the tracker’s protein-design evidence into experimentally validated small-molecule recognition, a harder interface that is relevant to drug delivery, sensing, and catalytic design.
Caveat
The study targets two selected small molecules under controlled assays; binding performance does not demonstrate general in-vivo delivery, sensing, or therapeutic use.