A general-purpose protein design framework based on mining sequence–structure relationships in known protein structures

By Jianfu Zhou, A. Panaitiu, Gevorg Grigoryan. 2018

89 citations recorded by the source corpus.

Significance Evolution has given us proteins that perform amazingly complex tasks in living systems, each molecule appearing “custom-built” for its particular purpose. Protein design seeks to enable the “custom building” of proteins at will, for specific tasks, without waiting for evolution. This is a grand challenge, because how a protein’s 3-dimensional structure and function are encoded in its amino acid sequence is exceedingly difficult to model. In this paper, we argue that sequence–structure encodings can instead be learned directly from proteins of known structure, which enables an approach to design. We are at an exciting time in protein science, where emergent principles inferred from data may allow us to make headway in cases where application of first principles is challenging. Current state-of-the-art approaches to computational protein design (CPD) aim to capture the determinants of structure from physical principles. While this has led to many successful designs, it does have strong limitations associated with inaccuracies in physical modeling, such that a reliable general solution to CPD has yet to be found. Here, we propose a design framework—one based on identifying and applying patterns of sequence–structure compatibility found in known proteins, rather than approximating them from models of interatomic interactions. We carry out extensive computational analyses and an experimental validation for our method. Our results strongly argue that the Protein Data Bank is now sufficiently large to enable proteins to be designed by using only examples of structural motifs from unrelated proteins. Because our method is likely to have orthogonal strengths relative to existing techniques, it could represent an important step toward removing remaining barriers to robust CPD.

Subfields: Materials Chemistry, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Materials Chemistry

In Molecular Biology

Citations in the corpus

Learning inverse folding from millions of predicted structures (2022, methodology): “An alternative non-parametric approach involves decomposing the library of known structures into common sequence-structure motifs (Zhou et al., 2020).”
Masked Inverse Folding with Sequence Transfer for Protein Representation Learning (2023, methodology): “Outside of deep-learning based methods, researchers use packing algorithms (Dahiyat & Mayo, 1997; Street & Mayo, 1999; DeGrado et al., 1991; Harbury et al., 1998), physics-based energy functions (Alford et al., 2017), or match structural motifs to sequence motifs (Zhou et al., 2020).”
Illuminating protein space with a programmable generative model (2022, background): “Natural protein structures exhibit considerable degeneracy in their use of local tertiary backbone geometries, such that completely unrelated proteins tend to use very similar tertiary motifs 37,38 .”
Perturbing the energy landscape for improved packing during computational protein design (2020, unknown): “4-12 Common solutions include reducing repulsive forces to implicitly account for small structural changes that may relieve clashes, allowing small localized changes to the protein backbone during sequence optimization, and iterating between sequence design and larger-scale structure refinement.”
Generative Models for Graph-Based Protein Design (2019, background): “More recently, there have been successes with non-parametric approaches to protein design [32] that are based on ﬁnding substructural homologies between the target and diverse templates in large protein database.”
Protein sequence design with a learned potential (2020, methodology): “…methods that account for ﬂexibility of the protein backbone often use “softer” potentials [44, 49], such as statistical potentials that are derived from data [50, 51, 50, 52, 53, 54], but these methods often do not perform as favorably as well-parameterized "hard" molecular mechanics force ﬁelds.”