Geometricus represents protein structures as shape-mers derived from moment invariants

Foldclass and Merizo-search: scalable structural similarity search for single- and multi-domain proteins using geometric learning

MOTIVATION

The availability of very large numbers of protein structures from accurate computational methods poses new challenges in storing, searching and detecting relationships between these structures. In particular, the new-found abundance of multi-domain structures in the AlphaFold structure database introduces challenges for traditional structure comparison methods.

RESULTS

We address these challenges using a fast, embedding-based structure comparison method called Foldclass which detects structural similarity between protein domains. We demonstrate the accuracy of Foldclass embeddings for homology detection. In combination with a recently developed deep learning-based automatic domain segmentation tool Merizo, we develop Merizo-search, which first segments multi-domain query structures into domains, and then searches a Foldclass embedding database to determine the top matches for each constituent domain. Combining the ability of Merizo to accurately segment complete chains into domains, and Foldclass to embed and detect similar domains, the Merizo-search tool can be used to rapidly detect per-domain similarities for complete chains, taking as little as 2 min to search all 365 million domains from the Encyclopedia of Domains. We anticipate that these tools will enable many analyses using the wealth of predicted structural data now available.

AVAILABILITY AND IMPLEMENTATION

Foldclass and Merizo-search are available at https://github.com/psipred/merizo_search. The version used in this publication is archived at https://doi.org/10.5281/zenodo.15120830. Merizo-search is also available on the PSIPRED web server at http://bioinf.cs.ucl.ac.uk/psipred.

Topology-based protein classification: A deep learning approach

Utilizing Artificial Intelligence (AI) in computational biology techniques could offer significant advantages in alleviating the growing workloads faced by structural biologists, especially with the emergence of big data. In this study, we employed Delaunay tessellation as a promising method to obtain the overall structural topology of proteins. Subsequently, we developed multi-class deep neural network models to classify protein superfamilies based on their local topology. Our models achieved a test accuracy of approximately 0.92 in classifying proteins into 18 well-populated superfamilies. We believe that the results of this study hold substantial value since, to the best of our knowledge, no previous studies have reported the utilization of protein topological data for protein classification through deep learning and Delaunay tessellation.

Deep generative models of protein structure uncover distant relationships across a continuous fold space

Our views of fold space implicitly rest upon many assumptions that impact how we analyze, interpret and understand protein structure, function and evolution. For instance, is there an optimal granularity in viewing protein structural similarities (e.g., architecture, topology or some other level)? Similarly, the discrete/continuous dichotomy of fold space is central, but remains unresolved. Discrete views of fold space bin similar folds into distinct, non-overlapping groups; unfortunately, such binning can miss remote relationships. While hierarchical systems like CATH are indispensable resources, less heuristic and more conceptually flexible approaches could enable more nuanced explorations of fold space. Building upon an Urfold model of protein structure, here we present a deep generative modeling framework, termed DeepUrfold, for analyzing protein relationships at scale. DeepUrfold's learned embeddings occupy high-dimensional latent spaces that can be distilled for a given protein in terms of an amalgamated representation uniting sequence, structure and biophysical properties. This approach is structure-guided, versus being purely structure-based, and DeepUrfold learns representations that, in a sense, define superfamilies. Deploying DeepUrfold with CATH reveals evolutionarily-remote relationships that evade existing methodologies, and suggests a mostly-continuous view of fold space-a view that extends beyond simple geometric similarity, towards the realm of integrated sequence ↔ structure ↔ function properties.

iMAX FRET (Information Maximized FRET) for Multipoint Single-Molecule Structural Analysis

Understanding the structure of biomolecules is vital for deciphering their roles in biological systems. Single-molecule techniques have emerged as alternatives to conventional ensemble structure analysis methods for uncovering new biology in molecular dynamics and interaction studies, yet only limited structural information could be obtained experimentally. Here, we address this challenge by introducing iMAX FRET, a one-pot method that allows ab initio 3D profiling of individual molecules using two-color FRET measurements. Through the stochastic exchange of fluorescent weak binders, iMAX FRET simultaneously assesses multiple distances on a biomolecule within a few minutes, which can then be used to reconstruct the coordinates of up to four points in each molecule, allowing structure-based inference. We demonstrate the 3D reconstruction of DNA nanostructures, protein quaternary structures, and conformational changes in proteins. With iMAX FRET, we provide a powerful approach to advance the understanding of biomolecular structure by expanding conventional FRET analysis to three dimensions.

Utilizing the scale-invariant feature transform algorithm to align distance matrices facilitates systematic protein structure comparison

MOTIVATION

Protein structure comparison is pivotal for deriving homological relationships, elucidating protein functions, and understanding evolutionary developments. The burgeoning field of in-silico protein structure prediction now yields billions of models with near-experimental accuracy, necessitating sophisticated tools for discerning structural similarities among proteins, particularly when sequence similarity is limited.

RESULTS

In this article, we have developed the align distance matrix with scale (ADAMS) pipeline, which synergizes the distance matrix alignment method with the scale-invariant feature transform algorithm, streamlining protein structure comparison on a proteomic scale. Utilizing a computer vision-centric strategy for contrasting disparate distance matrices, ADAMS adeptly alleviates challenges associated with proteins characterized by a high degree of structural flexibility. Our findings indicate that ADAMS achieves a level of performance and accuracy on par with Foldseek, while maintaining similar speed. Crucially, ADAMS overcomes certain limitations of Foldseek in handling structurally flexible proteins, establishing it as an efficacious tool for in-depth protein structure analysis with heightened accuracy.

AVAILABILITY

ADAMS can be download and used as a python package from Python Package Index (PyPI): adams · PyPI. Source code and other materials are available from young55775/ADAMS-developing (github.com). An online server is available: Bseek Search Server (cryonet.ai).

Deep learning for protein structure prediction and design-progress and applications

Proteins are the key molecular machines that orchestrate all biological processes of the cell. Most proteins fold into three-dimensional shapes that are critical for their function. Studying the 3D shape of proteins can inform us of the mechanisms that underlie biological processes in living cells and can have practical applications in the study of disease mutations or the discovery of novel drug treatments. Here, we review the progress made in sequence-based prediction of protein structures with a focus on applications that go beyond the prediction of single monomer structures. This includes the application of deep learning methods for the prediction of structures of protein complexes, different conformations, the evolution of protein structures and the application of these methods to protein design. These developments create new opportunities for research that will have impact across many areas of biomedical research.

Protein interaction mapping reveals widespread targeting of development-related host transcription factors by phytoplasma effectors

Phytoplasmas are pathogenic bacteria that reprogram plant host development for their own benefit. Previous studies have characterized a few different phytoplasma effector proteins that destabilize specific plant transcription factors. However, these are only a small fraction of the potential effectors used by phytoplasmas; therefore, the molecular mechanisms through which phytoplasmas modulate their hosts require further investigation. To obtain further insights into the phytoplasma infection mechanisms, we generated a protein-protein interaction network between a broad set of phytoplasma effectors and a large, unbiased collection of Arabidopsis thaliana transcription factors and transcriptional regulators. We found widespread, but specific, interactions between phytoplasma effectors and host transcription factors, especially those related to host developmental processes. In particular, many unrelated effectors target specific sets of TCP transcription factors, which regulate plant development and immunity. Comparison with other host-pathogen protein interaction networks shows that phytoplasma effectors have unusual targets, indicating that phytoplasmas have evolved a unique and unusual infection strategy. This study contributes a rich and solid data source that guides further investigations of the functions of individual effectors, as demonstrated for some herein. Moreover, the dataset provides insights into the underlying molecular mechanisms of phytoplasma infection.

Fast and accurate protein structure search with Foldseek

As structure prediction methods are generating millions of publicly available protein structures, searching these databases is becoming a bottleneck. Foldseek aligns the structure of a query protein against a database by describing tertiary amino acid interactions within proteins as sequences over a structural alphabet. Foldseek decreases computation times by four to five orders of magnitude with 86%, 88% and 133% of the sensitivities of Dali, TM-align and CE, respectively.

Structure-guided metagenome mining to tap microbial functional diversity

Scientists now have access to millions of accurate three-dimensional (3D) models of protein structures. How do we leverage 3D structural models to learn about microbial functions encoded in metagenomes? Here, we review recent developments using protein structural features to mine metagenomes from diverse environments ranging from the human gut to soil and ocean viromes. We compare 3D protein structural methods to characterize antibiotic resistance phenotypes, nutrient cycling, and host-drug-microbe interactions. Broadly, we encourage the scientific community to look beyond global sequence and structure alignments by considering fine-grained descriptors such as distance to ligand, active site, and tertiary interactions between amino acid residues scaling to microbiomes. Finally, we highlight structure-inspired approaches to chart new areas of microbial protein-coding sequence space.

Artificial intelligence for natural product drug discovery

Developments in computational omics technologies have provided new means to access the hidden diversity of natural products, unearthing new potential for drug discovery. In parallel, artificial intelligence approaches such as machine learning have led to exciting developments in the computational drug design field, facilitating biological activity prediction and de novo drug design for molecular targets of interest. Here, we describe current and future synergies between these developments to effectively identify drug candidates from the plethora of molecules produced by nature. We also discuss how to address key challenges in realizing the potential of these synergies, such as the need for high-quality datasets to train deep learning algorithms and appropriate strategies for algorithm validation.

Uncovering new families and folds in the natural protein universe

We are now entering a new era in protein sequence and structure annotation, with hundreds of millions of predicted protein structures made available through the AlphaFold database1. These models cover nearly all proteins that are known, including those challenging to annotate for function or putative biological role using standard homology-based approaches. In this study, we examine the extent to which the AlphaFold database has structurally illuminated this 'dark matter' of the natural protein universe at high predicted accuracy. We further describe the protein diversity that these models cover as an annotated interactive sequence similarity network, accessible at https://uniprot3d.org/atlas/AFDB90v4 . By searching for novelties from sequence, structure and semantic perspectives, we uncovered the β-flower fold, added several protein families to Pfam database2 and experimentally demonstrated that one of these belongs to a new superfamily of translation-targeting toxin-antitoxin systems, TumE-TumA. This work underscores the value of large-scale efforts in identifying, annotating and prioritizing new protein families. By leveraging the recent deep learning revolution in protein bioinformatics, we can now shed light into uncharted areas of the protein universe at an unprecedented scale, paving the way to innovations in life sciences and biotechnology.

Novel machine learning approaches revolutionize protein knowledge

Breakthrough methods in machine learning (ML), protein structure prediction, and novel ultrafast structural aligners are revolutionizing structural biology. Obtaining accurate models of proteins and annotating their functions on a large scale is no longer limited by time and resources. The most recent method to be top ranked by the Critical Assessment of Structure Prediction (CASP) assessment, AlphaFold 2 (AF2), is capable of building structural models with an accuracy comparable to that of experimental structures. Annotations of 3D models are keeping pace with the deposition of the structures due to advancements in protein language models (pLMs) and structural aligners that help validate these transferred annotations. In this review we describe how recent developments in ML for protein science are making large-scale structural bioinformatics available to the general scientific community.

Beyond sequence: Structure-based machine learning

Recent breakthroughs in protein structure prediction demarcate the start of a new era in structural bioinformatics. Combined with various advances in experimental structure determination and the uninterrupted pace at which new structures are published, this promises an age in which protein structure information is as prevalent and ubiquitous as sequence. Machine learning in protein bioinformatics has been dominated by sequence-based methods, but this is now changing to make use of the deluge of rich structural information as input. Machine learning methods making use of structures are scattered across literature and cover a number of different applications and scopes; while some try to address questions and tasks within a single protein family, others aim to capture characteristics across all available proteins. In this review, we look at the variety of structure-based machine learning approaches, how structures can be used as input, and typical applications of these approaches in protein biology. We also discuss current challenges and opportunities in this all-important and increasingly popular field.

A structural biology community assessment of AlphaFold2 applications

Most proteins fold into 3D structures that determine how they function and orchestrate the biological processes of the cell. Recent developments in computational methods for protein structure predictions have reached the accuracy of experimentally determined models. Although this has been independently verified, the implementation of these methods across structural-biology applications remains to be tested. Here, we evaluate the use of AlphaFold2 (AF2) predictions in the study of characteristic structural elements; the impact of missense variants; function and ligand binding site predictions; modeling of interactions; and modeling of experimental structural data. For 11 proteomes, an average of 25% additional residues can be confidently modeled when compared with homology modeling, identifying structural features rarely seen in the Protein Data Bank. AF2-based predictions of protein disorder and complexes surpass dedicated tools, and AF2 models can be used across diverse applications equally well compared with experimentally determined structures, when the confidence metrics are critically considered. In summary, we find that these advances are likely to have a transformative impact in structural biology and broader life-science research.

Fast protein structure comparison through effective representation learning with contrastive graph neural networks

Protein structure alignment algorithms are often time-consuming, resulting in challenges for large-scale protein structure similarity-based retrieval. There is an urgent need for more efficient structure comparison approaches as the number of protein structures increases rapidly. In this paper, we propose an effective graph-based protein structure representation learning method, GraSR, for fast and accurate structure comparison. In GraSR, a graph is constructed based on the intra-residue distance derived from the tertiary structure. Then, deep graph neural networks (GNNs) with a short-cut connection learn graph representations of the tertiary structures under a contrastive learning framework. To further improve GraSR, a novel dynamic training data partition strategy and length-scaling cosine distance are introduced. We objectively evaluate our method GraSR on SCOPe v2.07 and a new released independent test set from PDB database with a designed comprehensive performance metric. Compared with other state-of-the-art methods, GraSR achieves about 7%-10% improvement on two benchmark datasets. GraSR is also much faster than alignment-based methods. We dig into the model and observe that the superiority of GraSR is mainly brought by the learned discriminative residue-level and global descriptors. The web-server and source code of GraSR are freely available at www.csbio.sjtu.edu.cn/bioinf/GraSR/ for academic use.

The size and shape of protein structures vary considerably. Accurate protein structure comparison usually relies on structure alignment algorithms. However, superimposing two protein structures is relatively time-consuming, which makes it inappropriate for large-scale protein structure retrieval. Alignment-free algorithms are proposed for efficient protein structure comparison over the last few decades. These algorithms first transform the coordinates of atoms in two proteins to fixed-length vectors. Then, the comparison can be done by measuring the distance or similarity between two vectors, which is much faster than alignment. In this study, we propose a novel protein structure representation method for efficient structure comparison. Compared with other state-of-the-art alignment-free methods, our method achieves better performance on both ranking and multi-class classification tasks due to the powerful representation ability of deep graph neural networks. We dig into the model and observe that the superiority of our method is mainly brought by the learned discriminative residue-level and global descriptors.

Guide tree optimization with genetic algorithm to improve multiple protein 3D-structure alignment

MOTIVATION

With the increasing availability of 3D-data, the focus of comparative bioinformatic analysis is shifting from protein sequence alignments toward more content-rich 3D-alignments. This raises the need for new ways to improve the accuracy of 3D-superimposition.

RESULTS

We proposed guide tree optimization with genetic algorithm (GA) as a universal tool to improve the alignment quality of multiple protein 3D-structures systematically. As a proof of concept, we implemented the suggested GA-based approach in popular Matt and Caretta multiple protein 3D-structure alignment (M3DSA) algorithms, leading to a statistically significant improvement of the TM-score quality indicator by up to 220-1523% on 'SABmark Superfamilies' (in 49-77% of cases) and 'SABmark Twilight' (in 59-80% of cases) datasets. The observed improvement in collections of distant homologies highlights the potentials of GA to optimize 3D-alignments of diverse protein superfamilies as one plausible tool to study the structure-function relationship.

AVAILABILITY AND IMPLEMENTATION

The source codes of patched gaCaretta and gaMatt programs are available open-access at https://github.com/n-canter/gamaps.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Bioinformatic analysis of subfamily-specific regions in 3D-structures of homologs to study functional diversity and conformational plasticity in protein superfamilies

Local 3D-structural differences in homologous proteins contribute to functional diversity observed in a superfamily, but so far received little attention as bioinformatic analysis was usually carried out at the level of amino acid sequences. We have developed Zebra3D - the first-of-its-kind bioinformatic software for systematic analysis of 3D-alignments of protein families using machine learning. The new tool identifies subfamily-specific regions (SSRs) - patterns of local 3D-structure (i.e. single residues, loops, or secondary structure fragments) that are spatially equivalent within families/subfamilies, but are different among them, and thus can be associated with functional diversity and function-related conformational plasticity. Bioinformatic analysis of protein superfamilies by Zebra3D can be used to study 3D-determinants of catalytic activity and specific accommodation of ligands, help to prepare focused libraries for directed evolution or assist development of chimeric enzymes with novel properties by exchange of equivalent regions between homologs, and to characterize plasticity in binding sites. A companion Mustguseal web-server is available to automatically construct a 3D-alignment of functionally diverse proteins, thus reducing the minimal input required to operate Zebra3D to a single PDB code. The Zebra3D + Mustguseal combined approach provides the opportunity to systematically explore the value of SSRs in superfamilies and to use this information for protein design and drug discovery. The software is available open-access at https://biokinet.belozersky.msu.ru/Zebra3D.

Comparative host-coronavirus protein interaction networks reveal pan-viral disease mechanisms

The COVID-19 pandemic, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is a grave threat to public health and the global economy. SARS-CoV-2 is closely related to the more lethal but less transmissible coronaviruses SARS-CoV-1 and Middle East respiratory syndrome coronavirus (MERS-CoV). Here, we have carried out comparative viral-human protein-protein interaction and viral protein localization analyses for all three viruses. Subsequent functional genetic screening identified host factors that functionally impinge on coronavirus proliferation, including Tom70, a mitochondrial chaperone protein that interacts with both SARS-CoV-1 and SARS-CoV-2 ORF9b, an interaction we structurally characterized using cryo-electron microscopy. Combining genetically validated host factors with both COVID-19 patient genetic data and medical billing records identified molecular mechanisms and potential drug treatments that merit further molecular and clinical study.