麻豆淫院

Recent advances in computational protein design have depended mainly on neural networks and machine learning to generate binders. However, the complexity of protein鈥損rotein interactions and the limitations of data-driven models constrain future progress.

A team of researchers from the Max Planck Institute for Biology T眉bingen, the University T眉bingen, and the University Hospital T眉bingen have now developed a training-free computational pipeline that uses complementary shape matching to guide the creation of protein binders. The paper is in the journal Advanced Science.

"Despite significant advances in computational protein design in recent years, designing protein binders from scratch remains challenging," explains the study's first author, Kateryna Maksymenko.

"To date, the most successful approaches deploy neural networks. Our goal was to develop a training-free pipeline for binder design. We wanted a design pipeline that not only enables the creation of site-specific binders but also deepens our understanding of protein folding and function."

From concept to practical application

Researchers developed a powerful computer-based tool that can create small proteins designed to precisely stick to and block specific parts of disease-related proteins. This makes it possible to design new drugs to treat cancer or immune disorders, for example, from scratch that target tricky spots on proteins.

The study showcases the successful application of the novel approach to design protein binders targeting two biologically important molecules: the interleukin-7 receptor alpha (IL-7R伪), which plays a critical role in immunity and leukemogenesis, and the vascular endothelial growth factor (VEGF), a key angiogenic molecule and a therapeutic target in diverse diseases.

The pipeline integrates rapid computational selection of shape-matching scaffolds from extensive protein databases with physics-based interface design and molecular dynamics simulations to rank promising binder candidates. The top designs were then experimentally validated, demonstrating strong binding affinity, high stability, and potent activity in vitro and in vivo.

The novel binders exhibited tight binding to their targets, as well as strong inhibition of their downstream signaling. This approach can easily be generalized to address other clinically-relevant targets.

A vision of the future of protein engineering

The new approach allows the identification of near-optimal binder templates from the human proteome, facilitating the humanization of designed therapies, which lowers their immunogenic risks.

This new design pipeline not only simplifies the process but also offers us deeper insights into the physical fundamentals of protein function. Moreover, this approach allows us to incorporate artificial amino acids into designed proteins. The researchers hope this work will inspire broader applications and accelerate therapeutic discovery.

The design tool also relies exclusively on first principles, making it applicable not only for designing genetically-encodable protein binders, but also synthetic proteins that incorporate artificial amino acids.

This work paves the way for robust, efficient, and interpretable protein binder design, potentially transforming drug development and molecular diagnostics.