The native state structures of globular proteins are stable and well-packed indicating that self-interactions are favored over protein-solvent interactions under folding conditions. We use this as a guiding principle to derive the geometry of the building blocks of protein structures, alpha-helices and strands assembled into beta-sheets, with no adjustable parameters, no amino acid sequence information, and no chemistry. There is an almost perfect fit between the dictates of mathematics and physics and the rules of quantum chemistry. Our theory establishes an energy landscape that channels protein evolution by providing sequence-independent platforms for elaborating sequence-dependent functional diversity. Our work highlights the vital role of discreteness in life and has implications for the creation of artificial life and on the nature of life elsewhere in the cosmos.
Concepts rooted in physics are becoming increasingly important in biology as we transition to an era in which quantitative descriptions of all processes from molecular to cellular level are needed. In this essay I discuss two unexpected findings of universal behavior, uncommon in biology, in the self-assembly of proteins and RNA. These findings, which are surprising, reveal that physics ideas applied to biological problems ranging from folding to gene expression to cellular movement and communication between cells might lead to discovery of universal principles operating in adoptable living sy…
Protein function and activity are a consequence of its three-dimensional structure. Single particle analysis of cryogenic electron micrographs has radically changed structural biology allowing atomic reconstruction of almost any type of proteins. While such an approach provides snapshots of three-dimensional structural information that can be correlated with function, the new frontier of protein structural biology is in the fourth dimension, time. Here we propose the use of liquid phase electron microscopy to expand structural biology into dynamic studies. We apply here single particle analysi…
A small fraction of all protein structures characterized so far are entangled. The challenge of understanding the properties of these knotted proteins, and the why and the how of their natural folding process, has been taken up in the past decade with different approaches, such as structural characterization, in vitro experiments, and simulations of protein models with varying levels of complexity. The simplest among these are the lattice Gō models, which belong to the class of structure-based models, i.e., models that are biased to the native structure by explicitly including structural data.…
Salesforce Research（Salesforce研究院）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
Transformer architectures have proven to learn useful representations for protein classification and generation tasks. However, these representations present challenges in interpretability. In this work, we demonstrate a set of methods for analyzing protein Transformer models through the lens of attention. We show that attention: (1) captures the folding structure of proteins, connecting amino acids that are far apart in the underlying sequence, but spatially close in the three-dimensional structure, (2) targets binding sites, a key functional component of proteins, and (3) focuses on progress…
Argonne National Laboratory（阿贡国家实验室）；University of Chicago（芝加哥大学）；University of Pittsburgh（匹兹堡大学）
We outline recent developments in artificial intelligence (AI) and machine learning (ML) techniques for integrative structural biology of intrinsically disordered proteins (IDP) ensembles. IDPs challenge the traditional protein structure-function paradigm by adapting their conformations in response to specific binding partners leading them to mediate diverse, and often complex cellular functions such as biological signaling, self organization and compartmentalization. Obtaining mechanistic insights into their function can therefore be challenging for traditional structural determination techni…
In this paper we will review various aspects of the biology of prions and focus on what is currently known about the mammalian PrP prion. Also we briefly describe the prions of yeast and other fungi. Prions are infectious proteins behaving like genes, i.e. proteins that not only contain genetic information in its tertiary structure, i.e. its shape, but are also able to transmit and replicate in a manner analogous to genes but through very different mechanisms. The term prion is derived from "proteinaceous infectious particle" and arose from the Prusiner hypothesis that the infectious agent of …
Three, interrelated biologically-relevant examples of biased random walks are presented: (1) A model for DNA melting, modelled as DNA unzipping, which provides a way to illustrate the role of the Boltzmann factor in a venue well-known to biology and pre-medical students; (2) the activity of helicase motor proteins in unzipping double-stranded DNA, for example, at the replication fork, which is an example of a Brownian ratchet; (3) force generation by actin polymerization, which is another Brownian ratchet, and for which the force and actin-concentration dependence of the velocity of actin poly…
Hepatitis C virus helicase is a molecular motor that splits nucleic acid duplex structures during viral replication, therefore representing a promising target for antiviral treatment. Hence, a detailed understanding of the mechanism by which it operates would facilitate the development of efficient drug-assisted therapies aiming to inhibit helicase activity. Despite extensive investigations performed in the past, a thorough understanding of the activity of this important protein was lacking since the underlying internal conformational motions could not be resolved. Here we review investigation…
Classical molecular dynamics (MD) simulations will be able to reach sampling in the second timescale within five years, producing petabytes of simulation data at current force field accuracy. Notwithstanding this, MD will still be in the regime of low-throughput, high-latency predictions with average accuracy. We envisage that machine learning (ML) will be able to solve both the accuracy and time-to-prediction problem by learning predictive models using expensive simulation data. The synergies between classical, quantum simulations and ML methods, such as artificial neural networks, have the p…
OpenProteinSet: Training data for structural biology at scale
Harvard University（哈佛大学）；Harvard Medical School（哈佛医学院）；Columbia University（哥伦比亚大学）；Prescient Design, Genentech（普雷赛特设计公司，基因泰克）；New York University（纽约大学）；Flatiron Institute（弗拉蒂伦研究所）
From Prediction to Simulation: AlphaFold 3 as a Differentiable Framework for Structural Biology
Department of Computer Engineering, Ma.C., Islamic Azad University（计算机工程系，伊斯兰教Azad大学）；Department of Biological Sciences and Technologies, Faculty of Basic Sciences, Islamic Azad University（生物科学与技术系，基础科学学院，伊斯兰教Azad大学）
While modern structural biology has provided us with a rich and diverse picture of membrane proteins, the biological function of membrane proteins is often influenced by the mechanical properties of the surrounding lipid bilayer. Here we explore the relation between the shape of membrane proteins and the cooperative function of membrane proteins induced by membrane-mediated elastic interactions. For the experimental model system of mechanosensitive ion channels we find that the sign and strength of elastic interactions depend on the protein shape, yielding distinct cooperative gating curves fo…
The fields of structural biology and soft matter have independently sought out fundamental principles to rationalize protein crystallization. Yet the conceptual differences and the limited overlap between the two disciplines have thus far prevented a comprehensive understanding of the phenomenon to emerge. We conduct a computational study of proteins from the rubredoxin family that bridges the two fields. Using atomistic simulations, we characterize their crystal contacts, and accordingly parameterize patchy particle models. Comparing the phase diagrams of these schematic models with experimen…
We propose an application of molecular information theory to analyze the folding of single domain proteins. We analyze results from various areas of protein science, such as sequence-based potentials, reduced amino acid alphabets, backbone configurational entropy, secondary structure content, residue burial layers, and mutational studies of protein stability changes. We found that the average information contained in the sequences of evolved proteins is very close to the average information needed to specify a fold ~2.2 $\pm$ 0.3 bits/(site operation). The effective alphabet size in evolved pr…
One of the main concerns of Anfinsen was to reveal the connection between the amino acid sequence and their biologically active conformation. This search gave rise to two crucial questions in structural biology, namely, why the proteins fold and how a sequence encodes its folding. As to the why, he proposes a plausible answer, namely, at a given milieu a protein folds into its functional form -- native state -- because such structure represents the lowest free-energy minimum among all feasible conformations -- the thermodynamic hypothesis or Anfinsen dogma. As to the how, this remains as an un…
Proteins are the common constituents of all living cells. They are molecular machines that interact with each other as well as with other cell products and carry out a dizzying array of functions with distinction. These interactions follow from their native state structures and therefore understanding sequence-structure relationships is of fundamental importance. What is quite remarkable about proteins is that their understanding necessarily straddles several disciplines. The importance of geometry in defining protein native state structure, the constraints placed on protein behavior by mathem…
It's been said that "mathematics is biology's next microscope, only better; biology is mathematics' next physics, only better". Here we aim for something even better. We try to combine mathematical physics and biology into a picoscope of life. For this we merge techniques which have been introduced and developed in modern mathematical physics, largely by Ludvig Faddeev to describe objects such as solitons and Higgs and to explain phenomena such as anomalies in gauge fields. We propose a synthesis that can help to resolve the protein folding problem, one of the most important conundrums in all …
Protein folding is a central challenge in computational biology, with important applications in molecular biology, drug discovery and catalyst design. As a hard combinatorial optimisation problem, it has been studied as a potential target problem for quantum annealing. Although several experimental implementations have been discussed in the literature, the computational scaling of these approaches has not been elucidated. In this article, we present a numerical study of quantum annealing applied to a large number of small peptide folding problems, aiming to infer useful insights for near-term …
Protein fold classification is a classic problem in structural biology and bioinformatics. We approach this problem using persistent homology. In particular, we use alpha shape filtrations to compare a topological representation of the data with a different representation that makes use of knot-theoretic ideas. We use the statistical method of Angle-based Joint and Individual Variation Explained (AJIVE) to understand similarities and differences between these representations.
Protein function and dynamics are closely related to its sequence and structure. However prediction of protein function and dynamics from its sequence and structure is still a fundamental challenge in molecular biology. Protein classification, which is typically done through measuring the similarity be- tween proteins based on protein sequence or physical information, serves as a crucial step toward the understanding of protein function and dynamics. Persistent homology is a new branch of algebraic topology that has found its success in the topological data analysis in a variety of disciplines…
Despite an explosion in the number of experimentally determined, atomically detailed structures of biomolecules, many critical tasks in structural biology remain data-limited. Whether performance in such tasks can be improved by using large repositories of tangentially related structural data remains an open question. To address this question, we focused on a central problem in biology: predicting how proteins interact with one another---that is, which surfaces of one protein bind to those of another protein. We built a training dataset, the Database of Interacting Protein Structures (DIPS), t…
Antibodies are affinity proteins with a wide spectrum of applications in analytical and therapeutic biology. Proteins showing specific recognition for a chosen molecular target can be isolated and their encoding sequence identified in vitro from a large and diverse library by phage display selection. In this work, we show that this standard biochemical technique rapidly yields a collection of antibody protein binders for an inorganic target of major technological importance: crystalline metallic gold surfaces. 21 distinct anti-gold antibody proteins emerged from a large random library of antib…
As an example of topic where biology and physics meet, we present the issue of protein folding and stability, and the development of thermodynamics-based bioinformatics tools that predict the stability and thermal resistance of proteins and the change of these quantities upon amino acid substitutions. These methods are based on knowledge-driven statistical potentials, derived from experimental protein structures using the inverse Boltzmann law. We also describe an application of these predictors, which contributed to the understanding of the mechanisms of aggregation of a particular protein kn…
The key finding in the DNA double helix model is the specific pairing or binding between nucleotides A-T and C-G, and the pairing rules are the molecule basis of genetic code. Unfortunately, no such rules have been discovered for proteins. Here we show that similar rules and intrinsic sequence patterns between intra-protein binding peptide fragments do exist, and they can be extracted using a deep learning algorithm. Multi-millions of binding and non-binding peptide fragments from currently available protein X-ray structures are classified with an accuracy of up to 93%. This discovery has the …
Identifying and characterizing mutational paths is an important issue in evolutionary biology and in bioengineering. We here introduce a generic description of mutational paths in terms of the goodness of sequences and of the mutational dynamics (how sequences change) along the path. We first propose an algorithm to sample mutational paths, which we benchmark on exactly solvable models of proteins in silico, and apply to data-driven models of natural proteins learned from sequence data with Restricted Boltzmann Machines. We then use mean-field theory to characterize the properties of mutationa…
Seoul National University（首尔国立大学）；LG AI Research（LG人工智能研究院）；NAVER Corp.（NAVER公司）；Kangwon National University（江原国立大学）；Seoul National University of Science and Technology（首尔科技大学）
Bridging the exponentially growing gap between the numbers of unlabeled and labeled protein sequences, several studies adopted semi-supervised learning for protein sequence modeling. In these studies, models were pre-trained with a substantial amount of unlabeled data, and the representations were transferred to various downstream tasks. Most pre-training methods solely rely on language modeling and often exhibit limited performance. In this paper, we introduce a novel pre-training scheme called PLUS, which stands for Protein sequence representations Learned Using Structural information. PLUS …
This article introduces the physics of information in the context of molecular biology and genomics. Entropy and information, the two central concepts of Shannon's theory of information and communication, are often confused with each other but play transparent roles when applied to statistical ensembles (i.e., identically prepared sets) of symbolic sequences. Such an approach can distinguish between entropy and information in genes, predict the secondary structure of ribozymes, and detect the covariation between residues in folded proteins. We also review applications to molecular sequence and…
Generative modeling for protein engineering is key to solving fundamental problems in synthetic biology, medicine, and material science. We pose protein engineering as an unsupervised sequence generation problem in order to leverage the exponentially growing set of proteins that lack costly, structural annotations. We train a 1.2B-parameter language model, ProGen, on ~280M protein sequences conditioned on taxonomic and keyword tags such as molecular function and cellular component. This provides ProGen with an unprecedented range of evolutionary sequence diversity and allows it to generate wit…
Predicting the three-dimensional (3D) functional structures of proteins remains an important computational milestone in molecular biology to be achieved. This feat is hinged on a clear understanding of the mechanism which proteins use to fold into their native structures. Since Levinthal's paradox, there has been a lot of progress in understanding this mechanism. Most of the earlier attempts were caught between assigning either hydrophobic interactions or hydrogen bonding as the dominant folding force. However, a consensus now seems to be emerging about hydrogen bonding being a stronger force.…
The tertiary structure of protein, as well as the local secondary structure organization are fully determined by the angles of the peptidic bound. The backbone dihedral angles not only determine the global fold of the protein, but also the details of the local chain organization. Although a wealth of structural information is available in different databases and numerous structural biology softwares have been developed, rapid conformational characterization remains challenging. We present here RamaDA, a program able to give a synthetic description of the conformation of a protein. The RamaDA p…
The design of novel proteins has many applications but remains an attritional process with success in isolated cases. Meanwhile, deep learning technologies have exploded in popularity in recent years and are increasingly applicable to biology due to the rise in available data. We attempt to link protein design and deep learning by using variational autoencoders to generate protein sequences conditioned on desired properties. Potential copper and calcium binding sites are added to non-metal binding proteins without human intervention and compared to a hidden Markov model. In another use case, a…
Protein-protein interactions are fundamental to many biological processes. Experimental screens have identified tens of thousands of interactions and structural biology has provided detailed functional insight for select 3D protein complexes. An alternative rich source of information about protein interactions is the evolutionary sequence record. Building on earlier work, we show that analysis of correlated evolutionary sequence changes across proteins identifies residues that are close in space with sufficient accuracy to determine the three-dimensional structure of the protein complexes. We …
Concordia University（康考迪亚大学）；Rutgers University – Camden（罗格斯大学卡姆登分校）
Predicting the structure of a protein from its sequence is a cornerstone task of molecular biology. Established methods in the field, such as homology modeling and fragment assembly, appeared to have reached their limit. However, this year saw the emergence of promising new approaches: end-to-end protein structure and dynamics models, as well as reinforcement learning applied to protein folding. For these approaches to be investigated on a larger scale, an efficient implementation of their key computational primitives is required. In this paper we present a library of differentiable mappings f…
Information on protein-protein interactions (PPIs) not only advances our understanding of molecular biology but also provides important clues for target selection in drug discovery and the design of PPI inhibitors. One of the techniques used for computational prediction of PPIs is protein-protein docking calculations, and a variety of software has been developed. However, a friendly interface for users who are not sufficiently familiar with the command line interface has not been developed so far. In this study, we have developed a graphical user interface, MEGADOCK-GUI, which enables users to…
Mapping between sequence and structure is currently an open problem in structural biology. Despite many experimental and computational efforts it is not clear yet how the structure is encoded in the sequence. Answering this question may pave the way for predicting a protein fold given its sequence. My doctoral studies have focused on a particular phenomenon relevant to the protein sequence-structure relationship. It has been observed that many proteins having apparently dissimilar sequences share the same native fold. The phenomenon of mapping many divergent sequences into a single fold raises…
Background:Prediction of protein three-dimensional structures from amino acid sequences is a long-standing goal in computational/molecular biology. The successful discrimination of protein folds would help to improve the accuracy of protein 3D structure prediction. Results: In this work, we propose a method based on linear discriminant analysis (LDA) for recognizing proteins belonging to 30 different folds using the occurrence of amino acid residues in a set of 1612 proteins. The present method could discriminate the globular proteins from 30 major folding types with the sensitivity of 37%, wh…
Understanding the biological function of knots in proteins and their folding process is an open and challenging question in biology. Recent studies classify the topology and geometry of knotted proteins by analysing the distribution of a protein's planar projections using topological objects called knotoids. We approach the analysis of proteins with the same topology by introducing a topologically inspired statistical metric between their knotoid distributions. We detect geometric differences between trefoil proteins by characterising their entanglement and we recover a clustering by sequence …
Identifying novel functional protein structures is at the heart of molecular engineering and molecular biology, requiring an often computationally exhaustive search. We introduce the use of a Deep Convolutional Generative Adversarial Network (DCGAN) to classify protein structures based on their functionality by encoding each sample in a grid object structure using three features in each object: the generic atom type, the position atom type, and its occupancy relative to a given atom. We train DCGAN on 3-dimensional (3D) decoy and native protein structures in order to generate and discriminate …
Achieving a comprehensive understanding of the behaviour of proteins is greatly facilitated by the knowledge of their structures, thermodynamics and dynamics. All this information can be provided in an effective manner in terms of structural ensembles. A structural ensemble can be obtained by determining the structures, populations and interconversion rates for all the main states that a protein can occupy. To achieve this goal, integrative methods that combine experimental and computational approaches provide powerful tools. Here we focus on cryo-electron microscopy, which has become over rec…
Predicting protein structure from amino acid sequence is one of the most important unsolved problems of molecular biology and biophysics.Not only would a successful prediction algorithm be a tremendous advance in the understanding of the biochemical mechanisms of proteins, but, since such an algorithm could conceivably be used to design proteins to carry out specific functions.Prediction of the secondary structure of a protein (alpha-helix, beta-sheet, coil) is an important step towards elucidating its three dimensional structure as well as its function. In this research, we use different Hidd…
University of Notre Dame（圣母大学）；Eck Institute for Global Health, University of Notre Dame（圣母大学埃克全球健康研究所）；Interdisciplinary Center for Network Science & Applications, University of Notre Dame（圣母大学网络科学与应用跨学科中心）；University of Tennessee at Knoxville（田纳西大学诺克斯维尔分校）
As proteins with similar structures often have similar functions, analysis of protein structures can help predict protein functions and is thus important. We consider the problem of protein structure classification, which computationally classifies the structures of proteins into pre-defined groups. We develop a weighted network that depicts the protein structures, and more importantly, we propose the first graphlet-based measure that applies to weighted networks. Further, we develop a deep neural network (DNN) composed of both convolutional and recurrent layers to use this measure for classif…
Tsinghua University（清华大学）；Beijing Academy of Artificial Intelligence（北京智源人工智能研究院）；Tencent Quantum Lab（腾讯量子实验室）
Protein is linked to almost every life process. Therefore, analyzing the biological structure and property of protein sequences is critical to the exploration of life, as well as disease detection and drug discovery. Traditional protein analysis methods tend to be labor-intensive and time-consuming. The emergence of deep learning models makes modeling data patterns in large quantities of data possible. Interdisciplinary researchers have begun to leverage deep learning methods to model large biological datasets, e.g. using long short-term memory and convolutional neural network for protein sequ…
