Zero-shot design of drug-binding proteins via neural iterative selection-expansion

The design of proteins that bind to small molecules has been challenging due to the need for simultaneous optimization of the protein sequence, protein structure, and ligand conformation. Current deep-learning algorithms have struggled to navigate this complex landscape, precluding the zero-shot design of binders. Researchers have now developed a novel approach that combines two neural networks in an iterative design algorithm to create small-molecule binding proteins with high accuracy. This method, termed neural iterative selection-expansion (NISE), has been shown to outperform a comparable design loop using a physics-based energy function. By leveraging the strengths of two neural networks, LASErMPNN and a structure predictor, NISE optimizes sequence-structure-ligand compatibility, resulting in proteins that specifically bind to two chemically distinct small-molecule drugs, exatecan and apixaban, with high success rates. The tightest NISE binders exhibit nanomolar-to-picomolar affinities, surpassing those of the next-leading method by significant margins. Furthermore, LASErMPNN has suggested amino-acid substitutions that improve the affinity of the tightest exatecan binder by a substantial factor without requiring any experimental input. These optimized binders have been shown to protect the labile lactone ring of exatecan from hydrolysis for extended periods. The authors' work presents a general recipe for using neural networks to automate the design of small-molecule binding proteins for applications in drug delivery, sensing, and catalysis. However, it is essential to note that these proteins are intended for research-use-only laboratory applications and should not be used for therapeutic purposes without further testing and validation.