Edit |Radish peel.
Peptide research is of great significance for drug development, disease detection, and environmental monitoring. Many peptide hormones form helices after binding to their receptors, and detection methods for their sensitivity may contribute to better clinical management of the disease.
De novo protein design now enables the generation of conjugates with high affinity and specificity for structured proteins. However, the design of interactions between proteins and short peptides with helical tendencies is an unsolved challenge.
Scientists at the Protein Design Institute at the University of Washington School of Medicine are working with AI-driven biotechnology to address these challenges. They used artificial intelligence software to create protein molecules that bind to a variety of challenging biomarkers, including human hormones, with extremely high affinity and specificity.
Notably, the scientists achieved the highest intensity of interaction ever between the computer-generated biomolecule and its target.
D**id Baker, professor of medical biochemistry at the University of Washington and a research fellow at the Howard Hughes Medical Institute, highlights the potential impact of the approach: The ability to generate novel proteins with such high binding affinity and specificity opens up the world of possibilities from new diseases** to advanced diagnostics. 」
The study, titled De Novo Design of High-Affinity Binders of Bioactive Helical Peptides, was published in Nature on December 18, 2023.
Peptide hormones, such as parathyroid hormone (PTH), neuropeptide Y (NPY), glucagon (GCG), and secret-stimulating hormone (SCT), which adopt a helical structure after binding to their receptors, are recognized biomarkers in clinical care and biomedical research, playing a key role in human biology.
Preetham Venkatesh of Baker Labs explains: There are many diseases that are difficult to ** today because it is very challenging to detect certain molecules in the body. As diagnostic tools, designed proteins may provide a more cost-effective alternative to antibodies. 」
Scientists are very interested in their sensitivity and specificity, which currently relies on resource-intensive antibodies to generate, can be difficult to produce with high affinity, and often have less than ideal stability and reproducibility. The loop-mediated interaction surface of the antibody is not particularly well-suited to the highly specific binding of extended helical peptides—almost all antipeptide antibodies bind to their targets in a non-helical conformation.
Peptide-binding proteins are challenging to design
The designed protein can be easily produced in E. coli with high yield, low cost, and very high stability, however, the design of proteins that bind helical peptides with high affinity and specificity remains a prominent challenge. The design of peptide-binding proteins is challenging for two reasons.
First, proteins designed to bind folded proteins, such as 50-65 residue mini-binders with ultra-stable picomolar affinity, have a shape suitable for binding rigid concave targets, but are not suitable for peptides that support elongation. Helical peptides can easily bind to coil-forming coil assemblies, and this principle has been used to design binding agents for calmodulin peptides, but due to the large amount of exposed hydrophobic surface, the coiled helix subunit often self-associates without binding partners, greatly reducing the effective target binding affinity.
Second, peptides have fewer interaction residues and are often partially or completely unstructured at the time of separation. Therefore, constructing a peptide to a specific conformation may incur an entropy cost, which can compromise the favorable association free energy. Progress has been made in designing peptides that bind to the extended chain structure and polyproline II conformation, using protein side chains to interact with peptide backbones; However, due to the large number of backbone-backbone hydrogen bonds inside the helical peptide, this interaction cannot occur on the helical peptide.
New solutions
A team led by Susana Vazquez-Torres, Preetham Venkatesh, and Phil Leung, members of the Baker lab, is working to create proteins that bind to glucagon, neuropeptide Y, parathyroid hormone, and other helical peptide targets. The team proposed a new approach using RFDIFFUSION, a generative model for creating new protein shapes, in combination with the sequence design tool ProteinMPNN.
Researchers can refine the input structure model by extending the RFDIFFUSION to allow the binder design to adapt to more flexible targets and refine the input structure model with continuous noise and denoising (partial diffusion), and the picomolar affinity binder can generate a helical peptide target with a refined design generated by other methods, or start entirely from a random noise distribution.
Until now, these have been binding proteins designed for the highest affinity of any protein or small molecule target, and are computationally generated directly without any experimental optimization. The RFDIFFfusion design enables the enrichment and subsequent detection of parathyroid hormone and glucagon by mass spectrometry and the construction of bioluminescence-based protein biosensors. The ability to design conjugators for conformationally variable targets and the ability to optimize native and engineered proteins by partial diffusion should have a wide range of uses.
We are witnessing an exciting era of protein design, with advanced AI tools accelerating improvements in protein activity. This breakthrough will redefine the future of biotechnology. Vazquez-Torres noted.
The team collaborated with the lab of Joseph Rogers at the University of Copenhagen and the lab of Andrew Hoofnagle at the University of Washington School of Medicine to conduct lab tests to validate their biodesign approach. Mass spectrometry was used to detect engineered proteins bound to peptides at low concentrations in human serum, demonstrating the potential for sensitive and accurate disease diagnosis.
In addition, these proteins retain their target-binding capacity despite harsh conditions such as high temperatures, which is a key attribute for practical applications.
To further demonstrate the potential of this method, the researchers integrated a high-affinity parathyroid hormone binder into a biosensor system and achieved a 21-fold increase in bioluminescence signal in samples containing the target hormone. This integration with diagnostic equipment demonstrates the direct practical application of AI-generated proteins.
*Link: Related Reports: