Surface-centered design of de novo site-specific protein binders. a, Schematic of fingerprint generation. Spatially protein binding sites are included as vector fingerprints. The protein surfaces decompose into overlapping radial spots, and the neural network trained on native interacting protein pairs learns to embed the fingerprints such that complementary fingerprints are located in a similar region of space. We show an illustration of a subsample of a fingerprint projected into space reduced to three dimensions. The green box highlights a region of complementary fingerprints. b, MaSIF seeds – a method for identifying novel association seeds. The target patch is determined by the MaSIF location based on the propensity to form buried interfaces. Using MaSIF-seed, fingerprint integrity between the target patch and all fingerprints in a large database (~402 million patches) is evaluated; Fingerprint pairs are arranged later. The upper corrections are aligned and rescued to enable a more accurate assessment of the primary candidates. c, Scaffold search, seed grafting and interface redesign. Selected seeds are transferred to protein scaffolds and the rest of the interface is redesigned using Rosetta. The best designs are selected and tested empirically. credit: nature (2023). DOI: 10.1038/s41586-023-05993-x
Researchers from the Swiss Institute of Bioinformatics, Lausanne, Switzerland, used a deep learning engineering tool that produces “fingerprints” of protein surfaces to describe critical geometric and chemical features of protein-protein interactions. In their paper, “De novo design of protein interactions with acquired surface fingerprints” is published in natureThe team reported that the putative “fingerprints” captured key aspects of molecular recognition and novel protein interactions. A research brief summarizing the team’s findings is published in the same issue of the journal.
Proteins are the physical machinery of biology, the cogs, gears, springs and valves that allow organic life to function. It’s the way cells work, how drugs interact with biological systems, and where most diseases interact with biological systems. By perfecting this mechanism, science can cure most diseases.
Machine learning, together with proteomics, genetic sequencing, and molecular biology, has accelerated large-scale protein research over the past decade to the point where we can predict virtually all existing functional structures, engineer new ones in any configuration, and synthesize any protein. It can be imagined with one very small protein. But an important caveat – linking sites.
The way proteins interact depends on two specific physical “lock and key” interactions based on surface chemistry and structure. There is an outer rim site and a buried site underneath. The buried site does the work of the protein, but to access it, a leading outer edge signal is required to open the structure to allow access.
As much as proteins have emerged in the past decade, the affinity properties of proteins have remained elusive. This is partly because proteins are tough and pH-dependent, with changeable edge surface chemistries and binding sites that are state and site specific.
In the current research, the team sought structural affinity between proteins by essentially ignoring everything but surface affinity. Throwing away information about general structure, function, and similar protein interactions with their targets, the team focused machine learning on the proteins’ surface interactions, geometrical and chemical patterns that determine the best chance for two molecules to interact, and then designed switches.
By calculating fingerprints from the protein’s molecular surfaces, the team was able to quickly and reliably identify which complementary surface fragments could occupy a specific target within the 402 million candidate surfaces.
Several de novo protein ligands were computationally designed to engage four protein targets: SARS-CoV-2 spike, PD-1, PD-L1 and CTLA-4. Many of the designs have been improved experimentally, while others have been created in a purely digital space. The results were highly accurate affinity predictions as machine learning-based binders successfully engaged their targets.
The authors state that their framework could “…open up possibilities in other important biotechnology areas such as drug design, biosensors or biomaterials as well as provide a means to study interaction networks in biological processes at systems levels.”
more information:
Pablo Gainza et al, De novo design of protein interactions with acquired surface fingerprints, nature (2023). DOI: 10.1038/s41586-023-05993-x
new computer-designed protein-protein interactions, nature (2023). DOI: 10.1038/d41586-023-01324-2
© 2023 Science X Network
the quote: Solving the Mystery of Protein Surface Interactions with Geometric Fingerprints (2023, May 1) Retrieved May 1, 2023 from https://phys.org/news/2023-05-mystery-protein-surface-interactions-geometric.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.
