GRAFENE: Graphlet-based alignment-free network approach integrates 3D structural and sequence (residue order) data to improve protein structural comparison

Protein remote homology detection and structural alignment using deep learning

Exploiting sequence-structure-function relationships in biotechnology requires improved methods for aligning proteins that have low sequence similarity to previously annotated proteins. We develop two deep learning methods to address this gap, TM-Vec and DeepBLAST. TM-Vec allows searching for structure-structure similarities in large sequence databases. It is trained to accurately predict TM-scores as a metric of structural similarity directly from sequence pairs without the need for intermediate computation or solution of structures. Once structurally similar proteins have been identified, DeepBLAST can structurally align proteins using only sequence information by identifying structurally homologous regions between proteins. It outperforms traditional sequence alignment methods and performs similarly to structure-based alignment methods. We show the merits of TM-Vec and DeepBLAST on a variety of datasets, including better identification of remotely homologous proteins compared with state-of-the-art sequence alignment and structure prediction methods.

Freeprotmap: waiting-free prediction method for protein distance map

BACKGROUND

Protein residue-residue distance maps are used for remote homology detection, protein information estimation, and protein structure research. However, existing prediction approaches are time-consuming, and hundreds of millions of proteins are discovered each year, necessitating the development of a rapid and reliable prediction method for protein residue-residue distances. Moreover, because many proteins lack known homologous sequences, a waiting-free and alignment-free deep learning method is needed.

RESULT

In this study, we propose a learning framework named FreeProtMap. In terms of protein representation processing, the proposed group pooling in FreeProtMap effectively mitigates issues arising from high-dimensional sparseness in protein representation. In terms of model structure, we have made several careful designs. Firstly, it is designed based on the locality of protein structures and triangular inequality distance constraints to improve prediction accuracy. Secondly, inference speed is improved by using additive attention and lightweight design. Besides, the generalization ability is improved by using bottlenecks and a neural network block named local microformer. As a result, FreeProtMap can predict protein residue-residue distances in tens of milliseconds and has higher precision than the best structure prediction method.

CONCLUSION

Several groups of comparative experiments and ablation experiments verify the effectiveness of the designs. The results demonstrate that FreeProtMap significantly outperforms other state-of-the-art methods in accurate protein residue-residue distance prediction, which is beneficial for lots of protein research works. It is worth mentioning that we could scan all proteins discovered each year based on FreeProtMap to find structurally similar proteins in a short time because the fact that the structure similarity calculation method based on distance maps is much less time-consuming than algorithms based on 3D structures.

Graph embedding and geometric deep learning relevance to network biology and structural chemistry

Graphs are used as a model of complex relationships among data in biological science since the advent of systems biology in the early 2000. In particular, graph data analysis and graph data mining play an important role in biology interaction networks, where recent techniques of artificial intelligence, usually employed in other type of networks (e.g., social, citations, and trademark networks) aim to implement various data mining tasks including classification, clustering, recommendation, anomaly detection, and link prediction. The commitment and efforts of artificial intelligence research in network biology are motivated by the fact that machine learning techniques are often prohibitively computational demanding, low parallelizable, and ultimately inapplicable, since biological network of realistic size is a large system, which is characterised by a high density of interactions and often with a non-linear dynamics and a non-Euclidean latent geometry. Currently, graph embedding emerges as the new learning paradigm that shifts the tasks of building complex models for classification, clustering, and link prediction to learning an informative representation of the graph data in a vector space so that many graph mining and learning tasks can be more easily performed by employing efficient non-iterative traditional models (e.g., a linear support vector machine for the classification task). The great potential of graph embedding is the main reason of the flourishing of studies in this area and, in particular, the artificial intelligence learning techniques. In this mini review, we give a comprehensive summary of the main graph embedding algorithms in light of the recent burgeoning interest in geometric deep learning.

Metrics for network comparison using egonet feature distributions

Identifying networks with similar characteristics in a given ensemble, or detecting pattern discontinuities in a temporal sequence of networks, are two examples of tasks that require an effective metric capable of quantifying network (dis)similarity. Here we propose a method based on a global portrait of graph properties built by processing local nodes features. More precisely, a set of dissimilarity measures is defined by elaborating the distributions, over the network, of a few egonet features, namely the degree, the clustering coefficient, and the egonet persistence. The method, which does not require the alignment of the two networks being compared, exploits the statistics of the three features to define one- or multi-dimensional distribution functions, which are then compared to define a distance between the networks. The effectiveness of the method is evaluated using a standard classification test, i.e., recognizing the graphs originating from the same synthetic model. Overall, the proposed distances have performances comparable to the best state-of-the-art techniques (graphlet-based methods) with similar computational requirements. Given its simplicity and flexibility, the method is proposed as a viable approach for network comparison tasks.

Sequence-structure-function relationships in the microbial protein universe

For the past half-century, structural biologists relied on the notion that similar protein sequences give rise to similar structures and functions. While this assumption has driven research to explore certain parts of the protein universe, it disregards spaces that don't rely on this assumption. Here we explore areas of the protein universe where similar protein functions can be achieved by different sequences and different structures. We predict ~200,000 structures for diverse protein sequences from 1,003 representative genomes across the microbial tree of life and annotate them functionally on a per-residue basis. Structure prediction is accomplished using the World Community Grid, a large-scale citizen science initiative. The resulting database of structural models is complementary to the AlphaFold database, with regards to domains of life as well as sequence diversity and sequence length. We identify 148 novel folds and describe examples where we map specific functions to structural motifs. We also show that the structural space is continuous and largely saturated, highlighting the need for a shift in focus across all branches of biology, from obtaining structures to putting them into context and from sequence-based to sequence-structure-function based meta-omics analyses.

Multi-layer sequential network analysis improves protein 3D structural classification

Protein structural classification (PSC) is a supervised problem of assigning proteins into pre-defined structural (e.g., CATH or SCOPe) classes based on the proteins' sequence or 3D structural features. We recently proposed PSC approaches that model protein 3D structures as protein structure networks (PSNs) and analyze PSN-based protein features, which performed better than or comparable to state-of-the-art sequence or other 3D structure-based PSC approaches. However, existing PSN-based PSC approaches model the whole 3D structure of a protein as a static (i.e., single-layer) PSN. Because folding of a protein is a dynamic process, where some parts (i.e., sub-structures) of a protein fold before others, modeling the 3D structure of a protein as a PSN that captures the sub-structures might further help improve the existing PSC performance. Here, we propose to model 3D structures of proteins as multi-layer sequential PSNs that approximate 3D sub-structures of proteins, with the hypothesis that this will improve upon the current state-of-the-art PSC approaches that are based on single-layer PSNs (and thus upon the existing state-of-the-art sequence and other 3D structural approaches). Indeed, we confirm this on 72 datasets spanning ~44 000 CATH and SCOPe protein domains.

Understanding structural variability in proteins using protein structural networks

Proteins perform their function by accessing a suitable conformer from the ensemble of available conformations. The conformational diversity of a chosen protein structure can be obtained by experimental methods under different conditions. A key issue is the accurate comparison of different conformations. A gold standard used for such a comparison is the root mean square deviation (RMSD) between the two structures. While extensive refinements of RMSD evaluation at the backbone level are available, a comprehensive framework including the side chain interaction is not well understood. Here we employ protein structure network (PSN) formalism, with the non-covalent interactions of side chain, explicitly treated. The PSNs thus constructed are compared through graph spectral method, which provides a comparison at the local and at the global structural level. In this work, PSNs of multiple crystal conformers of single-chain, single-domain proteins, are subject to pair-wise analysis to examine the dissimilarity in their network topologies and in order to determine the conformational diversity of their native structures. This information is utilized to classify the structural domains of proteins into different categories. It is observed that proteins typically tend to retain structure and interactions at the backbone level. However, some of them also depict variability in either their overall structure or only in their inter-residue connectivity at the sidechain level, or both. Variability of sub-networks based on solvent accessibility and secondary structure is studied. The types of specific interactions are found to contribute differently to structure variability. An ensemble analysis by computing the mathematical variance of edge-weights across multiple conformers provided information on the contribution to overall variability from each edge of the PSN. Interactions that are highly variable are identified and their impact on structure variability has been discussed with the help of a case study. The classification based on the present side-chain network-based studies provides a framework to correlate the structure-function relationships in protein structures.

•

Monomeric, single domain protein structures can exhibit non-rigid behaviour and be highly variable.

•

The comparison of protein structural networks can better discriminate conformations with similar backbones.

•

Specific interactions between solvent accessible and inaccessible residues are poorly preserved.

•

Network edge-variation offers insights on which interacting residues are likely to influence their dynamics and function.

•

These side-chain network-based studies provide a framework to correlate protein structure-function relationships.

Modeling multi-scale data via a network of networks

MOTIVATION

Prediction of node and graph labels are prominent network science tasks. Data analyzed in these tasks are sometimes related: entities represented by nodes in a higher-level (higher scale) network can themselves be modeled as networks at a lower level. We argue that systems involving such entities should be integrated with a 'network of networks' (NoNs) representation. Then, we ask whether entity label prediction using multi-level NoN data via our proposed approaches is more accurate than using each of single-level node and graph data alone, i.e. than traditional node label prediction on the higher-level network and graph label prediction on the lower-level networks. To obtain data, we develop the first synthetic NoN generator and construct a real biological NoN. We evaluate accuracy of considered approaches when predicting artificial labels from the synthetic NoNs and proteins' functions from the biological NoN.

RESULTS

For the synthetic NoNs, our NoN approaches outperform or are as good as node- and network-level ones depending on the NoN properties. For the biological NoN, our NoN approaches outperform the single-level approaches for just under half of the protein functions, and for 30% of the functions, only our NoN approaches make meaningful predictions, while node- and network-level ones achieve random accuracy. So, NoN-based data integration is important.

AVAILABILITY AND IMPLEMENTATION

The software and data are available at https://nd.edu/~cone/NoNs.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Centrality Measures in Residue Interaction Networks to Highlight Amino Acids in Protein–Protein Binding

Residue interaction networks (RINs) describe a protein structure as a network of interacting residues. Central nodes in these networks, identified by centrality analyses, highlight those residues that play a role in the structure and function of the protein. However, little is known about the capability of such analyses to identify residues involved in the formation of macromolecular complexes. Here, we performed six different centrality measures on the RINs generated from the complexes of the SKEMPI 2 database of changes in protein–protein binding upon mutation in order to evaluate the capability of each of these measures to identify major binding residues. The analyses were performed with and without the crystallographic water molecules, in addition to the protein residues. We also investigated the use of a weight factor based on the inter-residue distances to improve the detection of these residues. We show that for the identification of major binding residues, closeness, degree, and PageRank result in good precision, whereas betweenness, eigenvector, and residue centrality analyses give a higher sensitivity. Including water in the analysis improves the sensitivity of all measures without losing precision. Applying weights only slightly raises the sensitivity of eigenvector centrality analysis. We finally show that a combination of multiple centrality analyses is the optimal approach to identify residues that play a role in protein–protein interaction.

Network analysis of synonymous codon usage

MOTIVATION

Most amino acids are encoded by multiple synonymous codons, some of which are used more rarely than others. Analyses of positions of such rare codons in protein sequences revealed that rare codons can impact co-translational protein folding and that positions of some rare codons are evolutionarily conserved. Analyses of their positions in protein 3-dimensional structures, which are richer in biochemical information than sequences alone, might further explain the role of rare codons in protein folding.

RESULTS

We model protein structures as networks and use network centrality to measure the structural position of an amino acid. We first validate that amino acids buried within the structural core are network-central, and those on the surface are not. Then, we study potential differences between network centralities and thus structural positions of amino acids encoded by conserved rare, non-conserved rare and commonly used codons. We find that in 84% of proteins, the three codon categories occupy significantly different structural positions. We examine protein groups showing different codon centrality trends, i.e. different relationships between structural positions of the three codon categories. We see several cases of all proteins from our data with some structural or functional property being in the same group. Also, we see a case of all proteins in some group having the same property. Our work shows that codon usage is linked to the final protein structure and thus possibly to co-translational protein folding.

AVAILABILITY AND IMPLEMENTATION

https://nd.edu/∼cone/CodonUsage/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

DeeplyEssential: a deep neural network for predicting essential genes in microbes

Background

Essential genes are those genes that are critical for the survival of an organism. The prediction of essential genes in bacteria can provide targets for the design of novel antibiotic compounds or antimicrobial strategies.

Results

We propose a deep neural network for predicting essential genes in microbes. Our architecture called DeeplyEssential makes minimal assumptions about the input data (i.e., it only uses gene primary sequence and the corresponding protein sequence) to carry out the prediction thus maximizing its practical application compared to existing predictors that require structural or topological features which might not be readily available. We also expose and study a hidden performance bias that effected previous classifiers. Extensive results show that DeeplyEssential outperform existing classifiers that either employ down-sampling to balance the training set or use clustering to exclude multiple copies of orthologous genes.

Conclusion

Deep neural network architectures can efficiently predict whether a microbial gene is essential (or not) using only its sequence information.

Chromatin network markers of leukemia

MOTIVATION

The structure of chromatin impacts gene expression. Its alteration has been shown to coincide with the occurrence of cancer. A key challenge is in understanding the role of chromatin structure (CS) in cellular processes and its implications in diseases.

RESULTS

We propose a comparative pipeline to analyze CSs and apply it to study chronic lymphocytic leukemia (CLL). We model the chromatin of the affected and control cells as networks and analyze the network topology by state-of-the-art methods. Our results show that CSs are a rich source of new biological and functional information about DNA elements and cells that can complement protein-protein and co-expression data. Importantly, we show the existence of structural markers of cancer-related DNA elements in the chromatin. Surprisingly, CLL driver genes are characterized by specific local wiring patterns not only in the CS network of CLL cells, but also of healthy cells. This allows us to successfully predict new CLL-related DNA elements. Importantly, this shows that we can identify cancer-related DNA elements in other cancer types by investigating the CS network of the healthy cell of origin, a key new insight paving the road to new therapeutic strategies. This gives us an opportunity to exploit chromosome conformation data in healthy cells to predict new drivers.

AVAILABILITY AND IMPLEMENTATION

Our predicted CLL genes and RNAs are provided as a free resource to the community at https://life.bsc.es/iconbi/chromatin/index.html.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Network-based protein structural classification

Experimental determination of protein function is resource-consuming. As an alternative, computational prediction of protein function has received attention. In this context, protein structural classification (PSC) can help, by allowing for determining structural classes of currently unclassified proteins based on their features, and then relying on the fact that proteins with similar structures have similar functions. Existing PSC approaches rely on sequence-based or direct three-dimensional (3D) structure-based protein features. By contrast, we first model 3D structures of proteins as protein structure networks (PSNs). Then, we use network-based features for PSC. We propose the use of graphlets, state-of-the-art features in many research areas of network science, in the task of PSC. Moreover, because graphlets can deal only with unweighted PSNs, and because accounting for edge weights when constructing PSNs could improve PSC accuracy, we also propose a deep learning framework that automatically learns network features from weighted PSNs. When evaluated on a large set of approximately 9400 CATH and approximately 12 800 SCOP protein domains (spanning 36 PSN sets), the best of our proposed approaches are superior to existing PSC approaches in terms of accuracy, with comparable running times. Our data and code are available at https://doi.org/10.5281/zenodo.3787922.

Quantitative description and classification of protein structures by a novel robust amino acid network: interaction selective network (ISN)

To quantitatively categorize protein structures, we developed a quantitative coarse-grained model of protein structures with a novel amino acid network, the interaction selective network (ISN), characterized by the links based on interactions in both the main and side chains. We found that the ISN is a novel robust network model to show the higher classification probability in the plots of average vertex degree (k) versus average clustering coefficient (C), both of which are typical network parameters for protein structures, and successfully distinguished between “all-α” and “all-β” proteins. On the other hand, one of the typical conventional networks, the α-carbon network (CAN), was found to be less robust than the ISN, and another typical network, atomic distance network (ADN), failed to distinguish between these two protein structures. Considering that the links in the CAN and ADN are defined by the interactions only between the main chain atoms and by the distance of the closest atom pair between the two amino acid residues, respectively, we can conclude that reflecting structural information from both secondary and tertiary structures in the network parameters improves the quantitative evaluation and robustness in network models, resulting in a quantitative and more robust description of three-dimensional protein structures in the ISN.

Comparing methods for comparing networks

With the impressive growth of available data and the flexibility of network modelling, the problem of devising effective quantitative methods for the comparison of networks arises. Plenty of such methods have been designed to accomplish this task: most of them deal with undirected and unweighted networks only, but a few are capable of handling directed and/or weighted networks too, thus properly exploiting richer information. In this work, we contribute to the effort of comparing the different methods for comparing networks and providing a guide for the selection of an appropriate one. First, we review and classify a collection of network comparison methods, highlighting the criteria they are based on and their advantages and drawbacks. The set includes methods requiring known node-correspondence, such as DeltaCon and Cut Distance, as well as methods not requiring a priori known node-correspondence, such as alignment-based, graphlet-based, and spectral methods, and the recently proposed Portrait Divergence and NetLSD. We test the above methods on synthetic networks and we assess their usability and the meaningfulness of the results they provide. Finally, we apply the methods to two real-world datasets, the European Air Transportation Network and the FAO Trade Network, in order to discuss the results that can be drawn from this type of analysis.

Functional geometry of protein interactomes

MOTIVATION

Protein-protein interactions (PPIs) are usually modeled as networks. These networks have extensively been studied using graphlets, small induced subgraphs capturing the local wiring patterns around nodes in networks. They revealed that proteins involved in similar functions tend to be similarly wired. However, such simple models can only represent pairwise relationships and cannot fully capture the higher-order organization of protein interactomes, including protein complexes.

RESULTS

To model the multi-scale organization of these complex biological systems, we utilize simplicial complexes from computational geometry. The question is how to mine these new representations of protein interactomes to reveal additional biological information. To address this, we define simplets, a generalization of graphlets to simplicial complexes. By using simplets, we define a sensitive measure of similarity between simplicial complex representations that allows for clustering them according to their data types better than clustering them by using other state-of-the-art measures, e.g. spectral distance, or facet distribution distance. We model human and baker's yeast protein interactomes as simplicial complexes that capture PPIs and protein complexes as simplices. On these models, we show that our newly introduced simplet-based methods cluster proteins by function better than the clustering methods that use the standard PPI networks, uncovering the new underlying functional organization of the cell. We demonstrate the existence of the functional geometry in the protein interactome data and the superiority of our simplet-based methods to effectively mine for new biological information hidden in the complexity of the higher-order organization of protein interactomes.

AVAILABILITY AND IMPLEMENTATION

Codes and datasets are freely available at http://www0.cs.ucl.ac.uk/staff/natasa/Simplets/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

An alignment-free measure based on physicochemical properties of amino acids for protein sequence comparison

Sequence comparison is an important topic in bioinformatics. With the exponential increase of biological sequences, the traditional protein sequence comparison methods - the alignment methods become limited, so the alignment-free methods are widely proposed in the past two decades. In this paper, we considered not only the six typical physicochemical properties of amino acids, but also their frequency and positional distribution. A 51-dimensional vector was obtained to describe the protein sequence. We got a pairwise distance matrix by computing the standardized Euclidean distance, and discriminant analysis and phylogenetic analysis can be made. The results on the Influenza A virus and ND5 datasets indicate that our method is accurate and efficient for classifying proteins and inferring the phylogeny of species.