CombFunc: predicting protein function using heterogeneous data sources

A large-scale assessment of sequence database search tools for homology-based protein function prediction

Sequence database searches followed by homology-based function transfer form one of the oldest and most popular approaches for predicting protein functions, such as Gene Ontology (GO) terms. These searches are also a critical component in most state-of-the-art machine learning and deep learning-based protein function predictors. Although sequence search tools are the basis of homology-based protein function prediction, previous studies have scarcely explored how to select the optimal sequence search tools and configure their parameters to achieve the best function prediction. In this paper, we evaluate the effect of using different options from among popular search tools, as well as the impacts of search parameters, on protein function prediction. When predicting GO terms on a large benchmark dataset, we found that BLASTp and MMseqs2 consistently exceed the performance of other tools, including DIAMOND-one of the most popular tools for function prediction-under default search parameters. However, with the correct parameter settings, DIAMOND can perform comparably to BLASTp and MMseqs2 in function prediction. Additionally, we developed a new scoring function to derive GO prediction from homologous hits that consistently outperform previously proposed scoring functions. These findings enable the improvement of almost all protein function prediction algorithms with a few easily implementable changes in their sequence homolog-based component. This study emphasizes the critical role of search parameter settings in homology-based function transfer and should have an important contribution to the development of future protein function prediction algorithms.

Intelligent Protein Design and Molecular Characterization Techniques: A Comprehensive Review

In recent years, the widespread application of artificial intelligence algorithms in protein structure, function prediction, and de novo protein design has significantly accelerated the process of intelligent protein design and led to many noteworthy achievements. This advancement in protein intelligent design holds great potential to accelerate the development of new drugs, enhance the efficiency of biocatalysts, and even create entirely new biomaterials. Protein characterization is the key to the performance of intelligent protein design. However, there is no consensus on the most suitable characterization method for intelligent protein design tasks. This review describes the methods, characteristics, and representative applications of traditional descriptors, sequence-based and structure-based protein characterization. It discusses their advantages, disadvantages, and scope of application. It is hoped that this could help researchers to better understand the limitations and application scenarios of these methods, and provide valuable references for choosing appropriate protein characterization techniques for related research in the field, so as to better carry out protein research.

A systematic review of state-of-the-art strategies for machine learning-based protein function prediction

New drug discovery is inseparable from the discovery of drug targets, and the vast majority of the known targets are proteins. At the same time, proteins are essential structural and functional elements of living cells necessary for the maintenance of all forms of life. Therefore, protein functions have become the focus of many pharmacological and biological studies. Traditional experimental techniques are no longer adequate for rapidly growing annotation of protein sequences, and approaches to protein function prediction using computational methods have emerged and flourished. A significant trend has been to use machine learning to achieve this goal. In this review, approaches to protein function prediction based on the sequence, structure, protein-protein interaction (PPI) networks, and fusion of multi-information sources are discussed. The current status of research on protein function prediction using machine learning is considered, and existing challenges and prominent breakthroughs are discussed to provide ideas and methods for future studies.

A Deep Learning Framework for Gene Ontology Annotations With Sequence- and Network-Based Information

Knowledge of protein functions plays an important role in biology and medicine. With the rapid development of high-throughput technologies, a huge number of proteins have been discovered. However, there are a great number of proteins without functional annotations. A protein usually has multiple functions and some functions or biological processes require interactions of a plurality of proteins. Additionally, Gene Ontology provides a useful classification for protein functions and contains more than 40,000 terms. We propose a deep learning framework called DeepGOA to predict protein functions with protein sequences and protein-protein interaction (PPI) networks. For protein sequences, we extract two types of information: sequence semantic information and subsequence-based features. We use the word2vec technique to numerically represent protein sequences, and utilize a Bi-directional Long and Short Time Memory (Bi-LSTM) and multi-scale convolutional neural network (multi-scale CNN) to obtain the global and local semantic features of protein sequences, respectively. Additionally, we use the InterPro tool to scan protein sequences for extracting subsequence-based information, such as domains and motifs. Then, the information is plugged into a neural network to generate high-quality features. For the PPI network, the Deepwalk algorithm is applied to generate its embedding information of PPI. Then the two types of features are concatenated together to predict protein functions. To evaluate the performance of DeepGOA, several different evaluation methods and metrics are utilized. The experimental results show that DeepGOA outperforms DeepGO and BLAST.

Structure-based protein function prediction using graph convolutional networks

The rapid increase in the number of proteins in sequence databases and the diversity of their functions challenge computational approaches for automated function prediction. Here, we introduce DeepFRI, a Graph Convolutional Network for predicting protein functions by leveraging sequence features extracted from a protein language model and protein structures. It outperforms current leading methods and sequence-based Convolutional Neural Networks and scales to the size of current sequence repositories. Augmenting the training set of experimental structures with homology models allows us to significantly expand the number of predictable functions. DeepFRI has significant de-noising capability, with only a minor drop in performance when experimental structures are replaced by protein models. Class activation mapping allows function predictions at an unprecedented resolution, allowing site-specific annotations at the residue-level in an automated manner. We show the utility and high performance of our method by annotating structures from the PDB and SWISS-MODEL, making several new confident function predictions. DeepFRI is available as a webserver at https://beta.deepfri.flatironinstitute.org/ .

Hierarchical Microbial Functions Prediction by Graph Aggregated Embedding

Matching 16S rRNA gene sequencing data to a metabolic reference database is a meaningful way to predict the metabolic function of bacteria and archaea, bringing greater insight to the working of the microbial community. However, some operational taxonomy units (OTUs) cannot be functionally profiled, especially for microbial communities from non-human samples cultured in defective media. Therefore, we herein report the development of Hierarchical micrObial functions Prediction by graph aggregated Embedding (HOPE), which utilizes co-occurring patterns and nucleotide sequences to predict microbial functions. HOPE integrates topological structures of microbial co-occurrence networks with k-mer compositions of OTU sequences and embeds them into a lower-dimensional continuous latent space, while maximally preserving topological relationships among OTUs. The high imbalance among KEGG Orthology (KO) functions of microbes is recognized in our framework that usually yields poor performance. A hierarchical multitask learning module is used in HOPE to alleviate the challenge brought by the long-tailed distribution among classes. To test the performance of HOPE, we compare it with HOPE-one, HOPE-seq, and GraphSAGE, respectively, in three microbial metagenomic 16s rRNA sequencing datasets, including abalone gut, human gut, and gut of Penaeus monodon. Experiments demonstrate that HOPE outperforms baselines on almost all indexes in all experiments. Furthermore, HOPE reveals significant generalization ability. HOPE's basic idea is suitable for other related scenarios, such as the prediction of gene function based on gene co-expression networks. The source code of HOPE is freely available at https://github.com/adrift00/HOPE.

NetGO: improving large-scale protein function prediction with massive network information

Automated function prediction (AFP) of proteins is of great significance in biology. AFP can be regarded as a problem of the large-scale multi-label classification where a protein can be associated with multiple gene ontology terms as its labels. Based on our GOLabeler—a state-of-the-art method for the third critical assessment of functional annotation (CAFA3), in this paper we propose NetGO, a web server that is able to further improve the performance of the large-scale AFP by incorporating massive protein-protein network information. Specifically, the advantages of NetGO are threefold in using network information: (i) NetGO relies on a powerful learning to rank framework from machine learning to effectively integrate both sequence and network information of proteins; (ii) NetGO uses the massive network information of all species (>2000) in STRING (other than only some specific species) and (iii) NetGO still can use network information to annotate a protein by homology transfer, even if it is not contained in STRING. Separating training and testing data with the same time-delayed settings of CAFA, we comprehensively examined the performance of NetGO. Experimental results have clearly demonstrated that NetGO significantly outperforms GOLabeler and other competing methods. The NetGO web server is freely available at http://issubmission.sjtu.edu.cn/netgo/.

Integrating multi-network topology for gene function prediction using deep neural networks

MOTIVATION

The emergence of abundant biological networks, which benefit from the development of advanced high-throughput techniques, contributes to describing and modeling complex internal interactions among biological entities such as genes and proteins. Multiple networks provide rich information for inferring the function of genes or proteins. To extract functional patterns of genes based on multiple heterogeneous networks, network embedding-based methods, aiming to capture non-linear and low-dimensional feature representation based on network biology, have recently achieved remarkable performance in gene function prediction. However, existing methods do not consider the shared information among different networks during the feature learning process.

RESULTS

Taking the correlation among the networks into account, we design a novel semi-supervised autoencoder method to integrate multiple networks and generate a low-dimensional feature representation. Then we utilize a convolutional neural network based on the integrated feature embedding to annotate unlabeled gene functions. We test our method on both yeast and human datasets and compare with three state-of-the-art methods. The results demonstrate the superior performance of our method. We not only provide a comprehensive analysis of the performance of the newly proposed algorithm but also provide a tool for extracting features of genes based on multiple networks, which can be used in the downstream machine learning task.

AVAILABILITY

DeepMNE-CNN is freely available at https://github.com/xuehansheng/DeepMNE-CNN.

CONTACT

jiajiepeng@nwpu.edu.cn; shang@nwpu.edu.cn; jianye.hao@tju.edu.cn.

Improving protein function prediction using protein sequence and GO-term similarities

MOTIVATION

Most automatic functional annotation methods assign Gene Ontology (GO) terms to proteins based on annotations of highly similar proteins. We advocate that proteins that are less similar are still informative. Also, despite their simplicity and structure, GO terms seem to be hard for computers to learn, in particular the Biological Process ontology, which has the most terms (>29 000). We propose to use Label-Space Dimensionality Reduction (LSDR) techniques to exploit the redundancy of GO terms and transform them into a more compact latent representation that is easier to predict.

RESULTS

We compare proteins using a sequence similarity profile (SSP) to a set of annotated training proteins. We introduce two new LSDR methods, one based on the structure of the GO, and one based on semantic similarity of terms. We show that these LSDR methods, as well as three existing ones, improve the Critical Assessment of Functional Annotation performance of several function prediction algorithms. Cross-validation experiments on Arabidopsis thaliana proteins pinpoint the superiority of our GO-aware LSDR over generic LSDR. Our experiments on A.thaliana proteins show that the SSP representation in combination with a kNN classifier outperforms state-of-the-art and baseline methods in terms of cross-validated F-measure.

AVAILABILITY AND IMPLEMENTATION

Source code for the experiments is available at https://github.com/stamakro/SSP-LSDR.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Exploring the computational methods for protein-ligand binding site prediction

Proteins participate in various essential processes in vivo via interactions with other molecules. Identifying the residues participating in these interactions not only provides biological insights for protein function studies but also has great significance for drug discoveries. Therefore, predicting protein-ligand binding sites has long been under intense research in the fields of bioinformatics and computer aided drug discovery. In this review, we first introduce the research background of predicting protein-ligand binding sites and then classify the methods into four categories, namely, 3D structure-based, template similarity-based, traditional machine learning-based and deep learning-based methods. We describe representative algorithms in each category and elaborate on machine learning and deep learning-based prediction methods in more detail. Finally, we discuss the trends and challenges of the current research such as molecular dynamics simulation based cryptic binding sites prediction, and highlight prospective directions for the near future.

The Genome3D Consortium for Structural Annotations of Selected Model Organisms

Genome3D consortium is a collaborative project involving protein structure prediction and annotation resources developed by six world-leading structural bioinformatics groups, based in the United Kingdom (namely Blundell, Murzin, Gough, Sternberg, Orengo, and Jones). The main objective of Genome3D serves as a common portal to provide both predicted models and annotations of proteins in model organisms, using several resources developed by these labs such as CATH-Gene3D, DOMSERF, pDomTHREADER, PHYRE, SUPERFAMILY, FUGUE/TOCATTA, and VIVACE. These resources primarily use SCOP- and/or CATH-based protein domain assignments. Another objective of Genome3D is to compare structural classifications of protein domains in CATH and SCOP databases and to provide a consensus mapping of CATH and SCOP protein superfamilies. CATH/SCOP mapping analyses led to the identification of total of 1429 consensus superfamilies.Currently, Genome3D provides structural annotations for ten model organisms, including Homo sapiens, Arabidopsis thaliana, Mus musculus, Escherichia coli, Saccharomyces cerevisiae, Caenorhabditis elegans, Drosophila melanogaster, Plasmodium falciparum, Staphylococcus aureus, and Schizosaccharomyces pombe. Thus, Genome3D serves as a common gateway to each structure prediction/annotation resource and allows users to perform comparative assessment of the predictions. It, thus, assists researchers to broaden their perspective on structure/function predictions of their query protein of interest in selected model organisms.

deepNF: deep network fusion for protein function prediction

Motivation

The prevalence of high-throughput experimental methods has resulted in an abundance of large-scale molecular and functional interaction networks. The connectivity of these networks provides a rich source of information for inferring functional annotations for genes and proteins. An important challenge has been to develop methods for combining these heterogeneous networks to extract useful protein feature representations for function prediction. Most of the existing approaches for network integration use shallow models that encounter difficulty in capturing complex and highly non-linear network structures. Thus, we propose deepNF, a network fusion method based on Multimodal Deep Autoencoders to extract high-level features of proteins from multiple heterogeneous interaction networks.

Results

We apply this method to combine STRING networks to construct a common low-dimensional representation containing high-level protein features. We use separate layers for different network types in the early stages of the multimodal autoencoder, later connecting all the layers into a single bottleneck layer from which we extract features to predict protein function. We compare the cross-validation and temporal holdout predictive performance of our method with state-of-the-art methods, including the recently proposed method Mashup. Our results show that our method outperforms previous methods for both human and yeast STRING networks. We also show substantial improvement in the performance of our method in predicting gene ontology terms of varying type and specificity.

Availability and implementation

deepNF is freely available at: https://github.com/VGligorijevic/deepNF.

Supplementary information

Supplementary data are available at Bioinformatics online.

Predicting Sequence Features, Function, and Structure of Proteins Using MESSA

MEta-Server for protein Sequence Analysis (MESSA) is a tool that facilitates widespread protein sequence analysis by gathering structural (local sequence properties and three-dimensional structure) and functional (annotations from SWISS-PROT, Gene Ontology terms, and enzyme classification) predictions for a query protein of interest. MESSA uses multiple well-established tools to offer consensus-based predictions on important aspects of protein sequence analysis. Being freely available for noncommercial users and with a user-friendly interface, MESSA serves as an umbrella platform that overcomes the absence of a comprehensive tool for predictive protein analysis. This article reveals how to access MESSA via the Web and shows how to input a protein sequence to analyze using the MESSA web server. It also includes a detailed explanation of the output from MESSA to aid in better interpretation of results. © 2019 by John Wiley & Sons, Inc.

UNIPred-Web: a web tool for the integration and visualization of biomolecular networks for protein function prediction

Background

One of the main issues in the automated protein function prediction (AFP) problem is the integration of multiple networked data sources. The UNIPred algorithm was thereby proposed to efficiently integrate —in a function-specific fashion— the protein networks by taking into account the imbalance that characterizes protein annotations, and to subsequently predict novel hypotheses about unannotated proteins. UNIPred is publicly available as R code, which might result of limited usage for non-expert users. Moreover, its application requires efforts in the acquisition and preparation of the networks to be integrated. Finally, the UNIPred source code does not handle the visualization of the resulting consensus network, whereas suitable views of the network topology are necessary to explore and interpret existing protein relationships.

Results

We address the aforementioned issues by proposing UNIPred-Web, a user-friendly Web tool for the application of the UNIPred algorithm to a variety of biomolecular networks, already supplied by the system, and for the visualization and exploration of protein networks. We support different organisms and different types of networks —e.g., co-expression, shared domains and physical interaction networks. Users are supported in the different phases of the process, ranging from the selection of the networks and the protein function to be predicted, to the navigation of the integrated network. The system also supports the upload of user-defined protein networks. The vertex-centric and the highly interactive approach of UNIPred-Web allow a narrow exploration of specific proteins, and an interactive analysis of large sub-networks with only a few mouse clicks.

Conclusions

UNIPred-Web offers a practical and intuitive (visual) guidance to biologists interested in gaining insights into protein biomolecular functions. UNIPred-Web provides facilities for the integration of networks, and supplies a framework for the imbalance-aware protein network integration of nine organisms, the prediction of thousands of GO protein functions, and a easy-to-use graphical interface for the visual analysis, navigation and interpretation of the integrated networks and of the functional predictions.

Environmental conditions shape the nature of a minimal bacterial genome

Of the 473 genes in the genome of the bacterium with the smallest genome generated to date, 149 genes have unknown function, emphasising a universal problem; less than 1% of proteins have experimentally determined annotations. Here, we combine the results from state-of-the-art in silico methods for functional annotation and assign functions to 66 of the 149 proteins. Proteins that are still not annotated lack orthologues, lack protein domains, and/ or are membrane proteins. Twenty-four likely transporter proteins are identified indicating the importance of nutrient uptake into and waste disposal out of the minimal bacterial cell in a nutrient-rich environment after removal of metabolic enzymes. Hence, the environment shapes the nature of a minimal genome. Our findings also show that the combination of multiple different state-of-the-art in silico methods for annotating proteins is able to predict functions, even for difficult to characterise proteins and identify crucial gaps for further development.

DeepFunc: A Deep Learning Framework for Accurate Prediction of Protein Functions from Protein Sequences and Interactions

Annotation of protein functions plays an important role in understanding life at the molecular level. High-throughput sequencing produces massive numbers of raw proteins sequences and only about 1% of them have been manually annotated with functions. Experimental annotations of functions are expensive, time-consuming and do not keep up with the rapid growth of the sequence numbers. This motivates the development of computational approaches that predict protein functions. A novel deep learning framework, DeepFunc, is proposed which accurately predicts protein functions from protein sequence- and network-derived information. More precisely, DeepFunc uses a long and sparse binary vector to encode information concerning domains, families, and motifs collected from the InterPro tool that is associated with the input protein sequence. This vector is processed with two neural layers to obtain a low-dimensional vector which is combined with topological information extracted from protein-protein interactions (PPIs) and functional linkages. The combined information is processed by a deep neural network that predicts protein functions. DeepFunc is empirically and comparatively tested on a benchmark testing dataset and the Critical Assessment of protein Function Annotation algorithms (CAFA) 3 dataset. The experimental results demonstrate that DeepFunc outperforms current methods on the testing dataset and that it secures the highest F_max  = 0.54 and AUC = 0.94 on the CAFA3 dataset.

DEEPred: Automated Protein Function Prediction with Multi-task Feed-forward Deep Neural Networks

Automated protein function prediction is critical for the annotation of uncharacterized protein sequences, where accurate prediction methods are still required. Recently, deep learning based methods have outperformed conventional algorithms in computer vision and natural language processing due to the prevention of overfitting and efficient training. Here, we propose DEEPred, a hierarchical stack of multi-task feed-forward deep neural networks, as a solution to Gene Ontology (GO) based protein function prediction. DEEPred was optimized through rigorous hyper-parameter tests, and benchmarked using three types of protein descriptors, training datasets with varying sizes and GO terms form different levels. Furthermore, in order to explore how training with larger but potentially noisy data would change the performance, electronically made GO annotations were also included in the training process. The overall predictive performance of DEEPred was assessed using CAFA2 and CAFA3 challenge datasets, in comparison with the state-of-the-art protein function prediction methods. Finally, we evaluated selected novel annotations produced by DEEPred with a literature-based case study considering the 'biofilm formation process' in Pseudomonas aeruginosa. This study reports that deep learning algorithms have significant potential in protein function prediction; particularly when the source data is large. The neural network architecture of DEEPred can also be applied to the prediction of the other types of ontological associations. The source code and all datasets used in this study are available at: https://github.com/cansyl/DEEPred .

Function Prediction for G Protein-Coupled Receptors through Text Mining and Induction Matrix Completion

G protein-coupled receptors (GPCRs) constitute the key component of cellular signal transduction. Accurately annotating the biological functions of GPCR proteins is vital to the understanding of the physiological processes they involve in. With the rapid development of text mining technologies and the exponential growth of biomedical literature, it becomes urgent to explore biological functional information from various literature for systematically and reliably annotating these known GPCRs. We design a novel three-stage approach, TM-IMC, using text mining and inductive matrix completion, for automated prediction of the gene ontology (GO) terms of the GPCR proteins. Large-scale benchmark tests show that inductive matrix completion models contribute to GPCR-GO association prediction for both molecular function and biological process aspects. Moreover, our detailed data analysis shows that information extracted from GPCR-associated literature indeed contributes to the prediction of GPCR-GO associations. The study demonstrated a new avenue to enhance the accuracy of GPCR function annotation through the combination of text mining and induction matrix completion over baseline methods in critical assessment of protein function annotation algorithms and literature-based GO annotation methods. Source codes of TM-IMC and the involved datasets can be freely downloaded from https://zhanglab.ccmb.med.umich.edu/TM-IMC for academic purposes.

DeepGO: predicting protein functions from sequence and interactions using a deep ontology-aware classifier

Motivation

A large number of protein sequences are becoming available through the application of novel high-throughput sequencing technologies. Experimental functional characterization of these proteins is time-consuming and expensive, and is often only done rigorously for few selected model organisms. Computational function prediction approaches have been suggested to fill this gap. The functions of proteins are classified using the Gene Ontology (GO), which contains over 40 000 classes. Additionally, proteins have multiple functions, making function prediction a large-scale, multi-class, multi-label problem.

Results

We have developed a novel method to predict protein function from sequence. We use deep learning to learn features from protein sequences as well as a cross-species protein–protein interaction network. Our approach specifically outputs information in the structure of the GO and utilizes the dependencies between GO classes as background information to construct a deep learning model. We evaluate our method using the standards established by the Computational Assessment of Function Annotation (CAFA) and demonstrate a significant improvement over baseline methods such as BLAST, in particular for predicting cellular locations.

Availability and implementation

Web server: http://deepgo.bio2vec.net, Source code: https://github.com/bio-ontology-research-group/deepgo

Supplementary information

Supplementary data are available at Bioinformatics online.

HPO2GO: prediction of human phenotype ontology term associations for proteins using cross ontology annotation co-occurrences

Analysing the relationships between biomolecules and the genetic diseases is a highly active area of research, where the aim is to identify the genes and their products that cause a particular disease due to functional changes originated from mutations. Biological ontologies are frequently employed in these studies, which provides researchers with extensive opportunities for knowledge discovery through computational data analysis. In this study, a novel approach is proposed for the identification of relationships between biomedical entities by automatically mapping phenotypic abnormality defining HPO terms with biomolecular function defining GO terms, where each association indicates the occurrence of the abnormality due to the loss of the biomolecular function expressed by the corresponding GO term. The proposed HPO2GO mappings were extracted by calculating the frequency of the co-annotations of the terms on the same genes/proteins, using already existing curated HPO and GO annotation sets. This was followed by the filtering of the unreliable mappings that could be observed due to chance, by statistical resampling of the co-occurrence similarity distributions. Furthermore, the biological relevance of the finalized mappings were discussed over selected cases, using the literature. The resulting HPO2GO mappings can be employed in different settings to predict and to analyse novel gene/protein—ontology term—disease relations. As an application of the proposed approach, HPO term—protein associations (i.e., HPO2protein) were predicted. In order to test the predictive performance of the method on a quantitative basis, and to compare it with the state-of-the-art, CAFA2 challenge HPO prediction target protein set was employed. The results of the benchmark indicated the potential of the proposed approach, as HPO2GO performance was among the best (Fmax = 0.35). The automated cross ontology mapping approach developed in this work may be extended to other ontologies as well, to identify unexplored relation patterns at the systemic level. The datasets, results and the source code of HPO2GO are available for download at: https://github.com/cansyl/HPO2GO.

Predicting human protein function with multi-task deep neural networks

Machine learning methods for protein function prediction are urgently needed, especially now that a substantial fraction of known sequences remains unannotated despite the extensive use of functional assignments based on sequence similarity. One major bottleneck supervised learning faces in protein function prediction is the structured, multi-label nature of the problem, because biological roles are represented by lists of terms from hierarchically organised controlled vocabularies such as the Gene Ontology. In this work, we build on recent developments in the area of deep learning and investigate the usefulness of multi-task deep neural networks (MTDNN), which consist of upstream shared layers upon which are stacked in parallel as many independent modules (additional hidden layers with their own output units) as the number of output GO terms (the tasks). MTDNN learns individual tasks partially using shared representations and partially from task-specific characteristics. When no close homologues with experimentally validated functions can be identified, MTDNN gives more accurate predictions than baseline methods based on annotation frequencies in public databases or homology transfers. More importantly, the results show that MTDNN binary classification accuracy is higher than alternative machine learning-based methods that do not exploit commonalities and differences among prediction tasks. Interestingly, compared with a single-task predictor, the performance improvement is not linearly correlated with the number of tasks in MTDNN, but medium size models provide more improvement in our case. One of advantages of MTDNN is that given a set of features, there is no requirement for MTDNN to have a bootstrap feature selection procedure as what traditional machine learning algorithms do. Overall, the results indicate that the proposed MTDNN algorithm improves the performance of protein function prediction. On the other hand, there is still large room for deep learning techniques to further enhance prediction ability.

MetaGO: Predicting Gene Ontology of Non-homologous Proteins Through Low-Resolution Protein Structure Prediction and Protein-Protein Network Mapping

Homology-based transferal remains the major approach to computational protein function annotations, but it becomes increasingly unreliable when the sequence identity between query and template decreases below 30%. We propose a novel pipeline, MetaGO, to deduce Gene Ontology attributes of proteins by combining sequence homology-based annotation with low-resolution structure prediction and comparison, and partner's homology-based protein-protein network mapping. The pipeline was tested on a large-scale set of 1000 non-redundant proteins from the CAFA3 experiment. Under the stringent benchmark conditions where templates with >30% sequence identity to the query are excluded, MetaGO achieves average F-measures of 0.487, 0.408, and 0.598, for Molecular Function, Biological Process, and Cellular Component, respectively, which are significantly higher than those achieved by other state-of-the-art function annotations methods. Detailed data analysis shows that the major advantage of the MetaGO lies in the new functional homolog detections from partner's homology-based network mapping and structure-based local and global structure alignments, the confidence scores of which can be optimally combined through logistic regression. These data demonstrate the power of using a hybrid model incorporating protein structure and interaction networks to deduce new functional insights beyond traditional sequence homology-based referrals, especially for proteins that lack homologous function templates. The MetaGO pipeline is available at http://zhanglab.ccmb.med.umich.edu/MetaGO/.

Assessing the Performances of Protein Function Prediction Algorithms from the Perspectives of Identification Accuracy and False Discovery Rate

The function of a protein is of great interest in the cutting-edge research of biological mechanisms, disease development and drug/target discovery. Besides experimental explorations, a variety of computational methods have been designed to predict protein function. Among these in silico methods, the prediction of BLAST is based on protein sequence similarity, while that of machine learning is also based on the sequence, but without the consideration of their similarity. This unique characteristic of machine learning makes it a good complement to BLAST and many other approaches in predicting the function of remotely relevant proteins and the homologous proteins of distinct function. However, the identification accuracies of these in silico methods and their false discovery rate have not yet been assessed so far, which greatly limits the usage of these algorithms. Herein, a comprehensive comparison of the performances among four popular prediction algorithms (BLAST, SVM, PNN and KNN) was conducted. In particular, the performance of these methods was systematically assessed by four standard statistical indexes based on the independent test datasets of 93 functional protein families defined by UniProtKB keywords. Moreover, the false discovery rates of these algorithms were evaluated by scanning the genomes of four representative model organisms (Homo sapiens, Arabidopsis thaliana, Saccharomyces cerevisiae and Mycobacterium tuberculosis). As a result, the substantially higher sensitivity of SVM and BLAST was observed compared with that of PNN and KNN. However, the machine learning algorithms (PNN, KNN and SVM) were found capable of substantially reducing the false discovery rate (SVM < PNN < KNN). In sum, this study comprehensively assessed the performance of four popular algorithms applied to protein function prediction, which could facilitate the selection of the most appropriate method in the related biomedical research.

Large-scale automated function prediction of protein sequences and an experimental case study validation on PTEN transcript variants

Recent advances in computing power and machine learning empower functional annotation of protein sequences and their transcript variations. Here, we present an automated prediction system UniGOPred, for GO annotations and a database of GO term predictions for proteomes of several organisms in UniProt Knowledgebase (UniProtKB). UniGOPred provides function predictions for 514 molecular function (MF), 2909 biological process (BP), and 438 cellular component (CC) GO terms for each protein sequence. UniGOPred covers nearly the whole functionality spectrum in Gene Ontology system and it can predict both generic and specific GO terms. UniGOPred was run on CAFA2 challenge target protein sequences and it is categorized within the top 10 best performing methods for the molecular function category. In addition, the performance of UniGOPred is higher compared to the baseline BLAST classifier in all categories of GO. UniGOPred predictions are compared with UniProtKB/TrEMBL database annotations as well. Furthermore, the proposed tool's ability to predict negatively associated GO terms that defines the functions that a protein does not possess, is discussed. UniGOPred annotations were also validated by case studies on PTEN protein variants experimentally and on CHD8 protein variants with literature. UniGOPred protein functional annotation system is available as an open access tool at http://cansyl.metu.edu.tr/UniGOPred.html.

Redox-Sensitive MarR Homologue BifR from Burkholderia thailandensis Regulates Biofilm Formation

Biofilm formation by pathogenic Burkholderia species is a serious complication as it renders the bacteria resistant to antibiotics and host defenses. Using B. thailandensis, we report here a novel redox-sensitive member of the multiple antibiotic resistance regulator (MarR) protein family, BifR, which represses biofilm formation. BifR is encoded as part of the emrB-bifR operon; emrB-bifR is divergent to ecsC, which encodes a putative LasA protease. In Pseudomonas aeruginosa, LasA has been implicated in virulence by contributing to cleavage of elastase. BifR repressed the expression of ecsC and emrB-bifR, and expression was further repressed under oxidizing conditions. BifR bound two sites in the intergenic region between ecsC and emrB-bifR with nanomolar affinity under both reducing and oxidizing conditions; however, oxidized BifR formed a disulfide-linked dimer-of-dimers, a covalent linkage that was absent in BifR-C104A in which the redox-active cysteine was replaced with alanine. BifR also repressed an operon encoding enzymes required for synthesis of phenazine antibiotics, which function as alternate respiratory electron receptors, and inactivation of bifR resulted in enhanced biofilm formation. Taken together, our data suggest that BifR functions to control LasA production and expression of genes involved in biofilm formation, in part by regulating synthesis of alternate electron acceptors that promote survival in the oxygen-limiting environment of a biofilm. The correlation between increased repression of emrB-bifR under oxidative conditions and the formation of a covalently linked BifR dimer-of-dimers suggests that BifR may modulate gene activity in response to cellular redox state.

Pipeline for inferring protein function from dynamics using coarse-grained molecular mechanics forcefield

Dynamics is integral to the function of proteins, yet the use of molecular dynamics (MD) simulation as a technique remains under-explored for molecular function inference. This is more important in the context of genomics projects where novel proteins are determined with limited evolutionary information. Recently we developed a method to match the query protein's flexible segments to infer function using a novel approach combining analysis of residue fluctuation-graphs and auto-correlation vectors derived from coarse-grained (CG) MD trajectory. The method was validated on a diverse dataset with sequence identity between proteins as low as 3%, with high function-recall rates. Here we share its implementation as a publicly accessible web service, named DynFunc (Dynamics Match for Function) to query protein function from ≥1 µs long CG dynamics trajectory information of protein subunits. Users are provided with the custom-developed coarse-grained molecular mechanics (CGMM) forcefield to generate the MD trajectories for their protein of interest. On upload of trajectory information, the DynFunc web server identifies specific flexible regions of the protein linked to putative molecular function. Our unique application does not use evolutionary information to infer molecular function from MD information and can, therefore, work for all proteins, including moonlighting and the novel ones, whenever structural information is available. Our pipeline is expected to be of utility to all structural biologists working with novel proteins and interested in moonlighting functions.

Inferring Function from Homology

Recent technological advances in sequencing and high-throughput DNA cloning have resulted in the generation of vast quantities of biological sequence data. Ideally the functions of individual genes and proteins predicted by these methods should be assessed experimentally within the context of a defined hypothesis. However, if no hypothesis is known a priori, or the number of sequences to be assessed is large, bioinformatics techniques may be useful in predicting function.This chapter proposes a pipeline of freely available Web-based tools to analyze protein-coding DNA and peptide sequences of unknown function. Accumulated information obtained during each step of the pipeline is used to build a testable hypothesis of function.The following methods are described in detail: 1. Annotation of gene function through Protein domain detection (SMART and Pfam). 2. Sequence similarity methods for homolog detection (BLAST and DELTA-BLAST). 3. Comparing sequences to whole genome data.

Computational Methods for Annotation Transfers from Sequence

Surveys of public sequence resources show that experimentally supported functional information is still completely missing for a considerable fraction of known proteins and is clearly incomplete for an even larger portion. Bioinformatics methods have long made use of very diverse data sources alone or in combination to predict protein function, with the understanding that different data types help elucidate complementary biological roles. This chapter focuses on methods accepting amino acid sequences as input and producing GO term assignments directly as outputs; the relevant biological and computational concepts are presented along with the advantages and limitations of individual approaches.

Protein Function Prediction

Protein function is a concept that can have different interpretations in different biological contexts, and the number and diversity of novel proteins identified by large-scale "omics" technologies poses increasingly new challenges. In this review we explore current strategies used to predict protein function focused on high-throughput sequence analysis, as for example, inference based on sequence similarity, sequence composition, structure, and protein-protein interaction. Various prediction strategies are discussed together with illustrative workflows highlighting the use of some benchmark tools and knowledge bases in the field.

Large-scale identification of human protein function using topological features of interaction network

The annotation of protein function is a vital step to elucidate the essence of life at a molecular level, and it is also meritorious in biomedical and pharmaceutical industry. Developments of sequencing technology result in constant expansion of the gap between the number of the known sequences and their functions. Therefore, it is indispensable to develop a computational method for the annotation of protein function. Herein, a novel method is proposed to identify protein function based on the weighted human protein-protein interaction network and graph theory. The network topology features with local and global information are presented to characterise proteins. The minimum redundancy maximum relevance algorithm is used to select 227 optimized feature subsets and support vector machine technique is utilized to build the prediction models. The performance of current method is assessed through 10-fold cross-validation test, and the range of accuracies is from 67.63% to 100%. Comparing with other annotation methods, the proposed way possesses a 50% improvement in the predictive accuracy. Generally, such network topology features provide insights into the relationship between protein functions and network architectures. The source code of Matlab is freely available on request from the authors.

SVM-Prot 2016: A Web-Server for Machine Learning Prediction of Protein Functional Families from Sequence Irrespective of Similarity

Knowledge of protein function is important for biological, medical and therapeutic studies, but many proteins are still unknown in function. There is a need for more improved functional prediction methods. Our SVM-Prot web-server employed a machine learning method for predicting protein functional families from protein sequences irrespective of similarity, which complemented those similarity-based and other methods in predicting diverse classes of proteins including the distantly-related proteins and homologous proteins of different functions. Since its publication in 2003, we made major improvements to SVM-Prot with (1) expanded coverage from 54 to 192 functional families, (2) more diverse protein descriptors protein representation, (3) improved predictive performances due to the use of more enriched training datasets and more variety of protein descriptors, (4) newly integrated BLAST analysis option for assessing proteins in the SVM-Prot predicted functional families that were similar in sequence to a query protein, and (5) newly added batch submission option for supporting the classification of multiple proteins. Moreover, 2 more machine learning approaches, K nearest neighbor and probabilistic neural networks, were added for facilitating collective assessment of protein functions by multiple methods. SVM-Prot can be accessed at http://bidd2.nus.edu.sg/cgi-bin/svmprot/svmprot.cgi.

Extensive complementarity between gene function prediction methods

MOTIVATION

The number of sequenced genomes rises steadily but we still lack the knowledge about the biological roles of many genes. Automated function prediction (AFP) is thus a necessity. We hypothesized that AFP approaches that draw on distinct genome features may be useful for predicting different types of gene functions, motivating a systematic analysis of the benefits gained by obtaining and integrating such predictions.

RESULTS

Our pipeline amalgamates 5 133 543 genes from 2071 genomes in a single massive analysis that evaluates five established genomic AFP methodologies. While 1227 Gene Ontology (GO) terms yielded reliable predictions, the majority of these functions were accessible to only one or two of the methods. Moreover, different methods tend to assign a GO term to non-overlapping sets of genes. Thus, inferences made by diverse genomic AFP methods display a striking complementary, both gene-wise and function-wise. Because of this, a viable integration strategy is to rely on a single most-confident prediction per gene/function, rather than enforcing agreement across multiple AFP methods. Using an information-theoretic approach, we estimate that current databases contain 29.2 bits/gene of known Escherichia coli gene functions. This can be increased by up to 5.5 bits/gene using individual AFP methods or by 11 additional bits/gene upon integration, thereby providing a highly-ranking predictor on the Critical Assessment of Function Annotation 2 community benchmark. Availability of more sequenced genomes boosts the predictive accuracy of AFP approaches and also the benefit from integrating them.

AVAILABILITY AND IMPLEMENTATION

The individual and integrated GO predictions for the complete set of genes are available from http://gorbi.irb.hr/ CONTACT: fran.supek@irb.hrSupplementary information: Supplementary materials are available at Bioinformatics online.

Streptococcus pneumoniae Genome-wide Identification and Characterization of BOX Element-binding Domains

The BOX elements are short repetitive DNA sequences that distribute randomly in intergenic regions of the Streptococcus pneumoniae genome. The function and origin of such elements are still unknown, but they were found to modulate expression of neighboring genes. Evidences suggested that the modulation's mechanism can be fulfilled by sequence-specific interaction of BOX elements with transcription factor family proteins. However, the type and function of these BOX-binding proteins still remain largely unexplored to date. In the current study we described a synthetic protocol to investigate the recognition and interaction between a highly conserved site of BOX elements and the DNA-binding domains of a variety of putative transcription factors in the pneumococcal genome. With the protocol we were able to predict those high-affinity domain binders of the conserved BOX DNA site (BOX DNA) in a high-throughput manner, and analyzed sequence-specific interaction in the domainDNA recognition at molecular level. Consequently, a number of putative transcription factor domains with both high affinity and specificity for the BOX DNA were identified, from which the helix-turn-helix (HTH) motif of a small heat shock factor was selected as a case study and tested for its binding capability toward the double-stranded BOX DNA using fluorescence anisotropy analysis. As might be expected, a relatively high affinity was detected for the interaction of HTH motif with BOX DNA with dissociation constant at nanomolar level. Molecular dynamics simulation, atomic structure examination and binding energy analysis revealed a complicated network of intensive nonbonded interactions across the complex interface, which confers both stability and specificity for the complex architecture.

Exploring Human Diseases and Biological Mechanisms by Protein Structure Prediction and Modeling

Protein structure prediction and modeling provide a tool for understanding protein functions by computationally constructing protein structures from amino acid sequences and analyzing them. With help from protein prediction tools and web servers, users can obtain the three-dimensional protein structure models and gain knowledge of functions from the proteins. In this chapter, we will provide several examples of such studies. As an example, structure modeling methods were used to investigate the relation between mutation-caused misfolding of protein and human diseases including epilepsy and leukemia. Protein structure prediction and modeling were also applied in nucleotide-gated channels and their interaction interfaces to investigate their roles in brain and heart cells. In molecular mechanism studies of plants, rice salinity tolerance mechanism was studied via structure modeling on crucial proteins identified by systems biology analysis; trait-associated protein-protein interactions were modeled, which sheds some light on the roles of mutations in soybean oil/protein content. In the age of precision medicine, we believe protein structure prediction and modeling will play more and more important roles in investigating biomedical mechanism of diseases and drug design.

Daily transcriptome changes reveal the role of nitrogen in controlling microcystin synthesis and nutrient transport in the toxic cyanobacterium, Microcystis aeruginosa

BACKGROUND

While transcriptomics have become a valuable tool for linking physiology and ecology in aquatic microbes, the temporal dynamics of global transcriptomic patterns in Microcystis have rarely been assessed. Furthermore, while many microbial studies have explored expression of nutrient transporter genes, few studies have concurrently measured nutrient assimilation rates. Here, we considered how the global transcriptomic patterns and physiology of the cyanobacterium, Microcystis aeruginosa, changed daily as cells were grown from replete to deficient nitrogen (N) conditions and then back to replete conditions.

RESULTS

During N deprivation, Microcystis downregulated genes involved in photosynthesis and respiration, carbon acquisition, lipid metabolism, and amino acid biosynthesis while upregulating genes involved in N acquisition and transport. With increasing N stress, both the strength of expression and number of genes being differentially expressed increased, until N was restored at which point these patterns reversed. Uptake of (15)N-labeled nitrate, ammonium and urea reflected differential expression of genes encoding transporters for these nutrients, with Microcystis appearing to preferentially increase transcription of ammonium and urea transporters and uptake of these compounds during N deprivation. Nitrate uptake and nitrate transporter expression were correlated for one set of transporters but not another, indicating these were high and low affinity nitrate transporters, respectively. Concentrations of microcystin per cell decreased during N deprivation and increased upon N restoration. However, the transcript abundance of genes involved in the synthesis of this compound was complex, as microcystin synthetase genes involved in peptide synthesis were downregulated under N deprivation while genes involved in tailoring and transport were upregulated, suggesting modification of the microcystin molecule under N stress as well as potential alternative functions for these genes and/or this toxin.

CONCLUSIONS

Collectively, this study highlights the complex choreography of gene expression, cell physiology, and toxin synthesis that dynamic N levels can elicit in this ecologically important cyanobacterium. Differing expression patterns of genes within the microcystin synthetase operon in response to changing N levels revealed the potential limitations drawing conclusions based on only one gene in this operon.

SIFTER search: a web server for accurate phylogeny-based protein function prediction

We are awash in proteins discovered through high-throughput sequencing projects. As only a minuscule fraction of these have been experimentally characterized, computational methods are widely used for automated annotation. Here, we introduce a user-friendly web interface for accurate protein function prediction using the SIFTER algorithm. SIFTER is a state-of-the-art sequence-based gene molecular function prediction algorithm that uses a statistical model of function evolution to incorporate annotations throughout the phylogenetic tree. Due to the resources needed by the SIFTER algorithm, running SIFTER locally is not trivial for most users, especially for large-scale problems. The SIFTER web server thus provides access to precomputed predictions on 16 863 537 proteins from 232 403 species. Users can explore SIFTER predictions with queries for proteins, species, functions, and homologs of sequences not in the precomputed prediction set. The SIFTER web server is accessible at http://sifter.berkeley.edu/ and the source code can be downloaded.

Genome-Wide Protein Function Prediction through Multi-Instance Multi-Label Learning

Automated annotation of protein function is challenging. As the number of sequenced genomes rapidly grows, the vast majority of proteins can only be annotated computationally. Nature often brings several domains together to form multi-domain and multi-functional proteins with a vast number of possibilities, and each domain may fulfill its own function independently or in a concerted manner with its neighbors. Thus, it is evident that the protein function prediction problem is naturally and inherently Multi-Instance Multi-Label (MIML) learning tasks. Based on the state-of-the-art MIML algorithm MIMLNN, we propose a novel ensemble MIML learning framework EnMIMLNN and design three algorithms for this task by combining the advantage of three kinds of Hausdorff distance metrics. Experiments on seven real-world organisms covering the biological three-domain system, i.e., archaea, bacteria, and eukaryote, show that the EnMIMLNN algorithms are superior to most state-of-the-art MIML and Multi-Label learning algorithms.

Structure to function prediction of hypothetical protein KPN_00953 (Ycbk) from Klebsiella pneumoniae MGH 78578 highlights possible role in cell wall metabolism

Background

Klebsiella pneumoniae plays a major role in causing nosocomial infection in immunocompromised patients. Medical inflictions by the pathogen can range from respiratory and urinary tract infections, septicemia and primarily, pneumonia. As more K. pneumoniae strains are becoming highly resistant to various antibiotics, treatment of this bacterium has been rendered more difficult. This situation, as a consequence, poses a threat to public health. Hence, identification of possible novel drug targets against this opportunistic pathogen need to be undertaken. In the complete genome sequence of K. pneumoniae MGH 78578, approximately one-fourth of the genome encodes for hypothetical proteins (HPs). Due to their low homology and relatedness to other known proteins, HPs may serve as potential, new drug targets.

Results

Sequence analysis on the HPs of K. pneumoniae MGH 78578 revealed that a particular HP termed KPN_00953 (YcbK) contains a M15_3 peptidases superfamily conserved domain. Some members of this superfamily are metalloproteases which are involved in cell wall metabolism. BLASTP similarity search on KPN_00953 (YcbK) revealed that majority of the hits were hypothetical proteins although two of the hits suggested that it may be a lipoprotein or related to twin-arginine translocation (Tat) pathway important for transport of proteins to the cell membrane and periplasmic space. As lipoproteins and other components of the cell wall are important pathogenic factors, homology modeling of KPN_00953 was attempted to predict the structure and function of this protein. Three-dimensional model of the protein showed that its secondary structure topology and active site are similar with those found among metalloproteases where two His residues, namely His169 and His209 and an Asp residue, Asp176 in KPN_00953 were found to be Zn-chelating residues. Interestingly, induced expression of the cloned KPN_00953 gene in lipoprotein-deficient E. coli JE5505 resulted in smoother cells with flattened edges. Some cells showed deposits of film-like material under scanning electron microscope.

Conclusions

We postulate that KPN_00953 is a Zn metalloprotease and may play a role in bacterial cell wall metabolism. Structural biology studies to understand its structure, function and mechanism of action pose the possibility of utilizing this protein as a new drug target against K. pneumoniae in the future.

Prediction of detailed enzyme functions and identification of specificity determining residues by random forests

Determining enzyme functions is essential for a thorough understanding of cellular processes. Although many prediction methods have been developed, it remains a significant challenge to predict enzyme functions at the fourth-digit level of the Enzyme Commission numbers. Functional specificity of enzymes often changes drastically by mutations of a small number of residues and therefore, information about these critical residues can potentially help discriminate detailed functions. However, because these residues must be identified by mutagenesis experiments, the available information is limited, and the lack of experimentally verified specificity determining residues (SDRs) has hindered the development of detailed function prediction methods and computational identification of SDRs. Here we present a novel method for predicting enzyme functions by random forests, EFPrf, along with a set of putative SDRs, the random forests derived SDRs (rf-SDRs). EFPrf consists of a set of binary predictors for enzymes in each CATH superfamily and the rf-SDRs are the residue positions corresponding to the most highly contributing attributes obtained from each predictor. EFPrf showed a precision of 0.98 and a recall of 0.89 in a cross-validated benchmark assessment. The rf-SDRs included many residues, whose importance for specificity had been validated experimentally. The analysis of the rf-SDRs revealed both a general tendency that functionally diverged superfamilies tend to include more active site residues in their rf-SDRs than in less diverged superfamilies, and superfamily-specific conservation patterns of each functional residue. EFPrf and the rf-SDRs will be an effective tool for annotating enzyme functions and for understanding how enzyme functions have diverged within each superfamily.

Going the distance for protein function prediction: a new distance metric for protein interaction networks

In protein-protein interaction (PPI) networks, functional similarity is often inferred based on the function of directly interacting proteins, or more generally, some notion of interaction network proximity among proteins in a local neighborhood. Prior methods typically measure proximity as the shortest-path distance in the network, but this has only a limited ability to capture fine-grained neighborhood distinctions, because most proteins are close to each other, and there are many ties in proximity. We introduce diffusion state distance (DSD), a new metric based on a graph diffusion property, designed to capture finer-grained distinctions in proximity for transfer of functional annotation in PPI networks. We present a tool that, when input a PPI network, will output the DSD distances between every pair of proteins. We show that replacing the shortest-path metric by DSD improves the performance of classical function prediction methods across the board.

MESSA: MEta-Server for protein Sequence Analysis

Background

Computational sequence analysis, that is, prediction of local sequence properties, homologs, spatial structure and function from the sequence of a protein, offers an efficient way to obtain needed information about proteins under study. Since reliable prediction is usually based on the consensus of many computer programs, meta-severs have been developed to fit such needs. Most meta-servers focus on one aspect of sequence analysis, while others incorporate more information, such as PredictProtein for local sequence feature predictions, SMART for domain architecture and sequence motif annotation, and GeneSilico for secondary and spatial structure prediction. However, as predictions of local sequence properties, three-dimensional structure and function are usually intertwined, it is beneficial to address them together.

Results

We developed a MEta-Server for protein Sequence Analysis (MESSA) to facilitate comprehensive protein sequence analysis and gather structural and functional predictions for a protein of interest. For an input sequence, the server exploits a number of select tools to predict local sequence properties, such as secondary structure, structurally disordered regions, coiled coils, signal peptides and transmembrane helices; detect homologous proteins and assign the query to a protein family; identify three-dimensional structure templates and generate structure models; and provide predictive statements about the protein's function, including functional annotations, Gene Ontology terms, enzyme classification and possible functionally associated proteins. We tested MESSA on the proteome of Candidatus Liberibacter asiaticus. Manual curation shows that three-dimensional structure models generated by MESSA covered around 75% of all the residues in this proteome and the function of 80% of all proteins could be predicted.

Availability

MESSA is free for non-commercial use at http://prodata.swmed.edu/MESSA/