Assessing the precision of high-throughput computational and laboratory approaches for the genome-wide identification of protein subcellular localization in bacteria

Protein Sorting Prediction

Many computational methods are available for predicting protein sorting in bacteria. When comparing them, it is important to know that they can be grouped into three fundamentally different approaches: signal-based, global property-based, and homology-based prediction. In this chapter, the strengths and drawbacks of each of these approaches are described through many examples of methods that predict secretion, integration into membranes, or subcellular locations in general. The aim of this chapter is to provide a user-level introduction to the field with a minimum of computational theory.

Proteomic profile of naturally released extracellular vesicles secreted from Leptospira interrogans serovar Pomona in response to temperature and osmotic stresses

Bacterial extracellular vesicles (EVs) are generally formed by pinching off outer membrane leaflets while simultaneously releasing multiple active molecules into the external environment. In this study, we aimed to identify the protein cargo of leptospiral EVs released from intact leptospires grown under three different conditions: EMJH medium at 30 °C, temperature shifted to 37 °C, and physiologic osmolarity (EMJH medium with 120 mM NaCl). The naturally released EVs observed under transmission electron microscopy were spherical in shape with an approximate diameter of 80-100 nm. Quantitative proteomics and bioinformatic analysis indicated that the EVs were formed primarily from the outer membrane and the cytoplasm. The main functional COG categories of proteins carried in leptospiral EVs might be involved in cell growth, survival and adaptation, and pathogenicity. Relative to their abundance in EVs grown in EMJH medium at 30 °C, 39 and 69 proteins exhibited significant changes in response to the temperature shift and the osmotic change, respectively. During exposure to both stresses, Leptospira secreted several multifunctional proteins via EVs, while preserving certain virulence proteins within whole cells. Therefore, leptospiral EVs may serve as a decoy structure for host responses, whereas some virulence factors necessary for direct interaction with the host environment are reserved in leptospiral cells. This knowledge will be useful for understanding the pathogenesis of leptospirosis and developing as one of vaccine platforms against leptospirosis in the future.

Pathogen-associated gene discovery workflows for novel antivirulence therapeutic development

Novel therapeutics to manage bacterial infections are urgently needed as the impact and prevalence of antimicrobial resistance (AMR) grows. Antivirulence therapeutics are an alternative approach to antibiotics that aim to attenuate virulence rather than target bacterial essential functions, while minimizing microbiota perturbation and the risk of AMR development. Beyond known virulence factors, pathogen-associated genes (PAGs; genes found only in pathogens to date) may play an important role in virulence or host association. Many identified PAGs encode uncharacterized hypothetical proteins and represent an untapped wealth of novel drug targets. Here, we review current advances in antivirulence drug research and development, including PAG identification, and provide a comprehensive workflow from the discovery of antivirulence drug targets to drug discovery. We highlight the importance of integrating bioinformatic/genomic-based methods for novel virulence factor discovery, coupled with experimental characterization, into existing drug screening platforms to develop novel and effective antivirulence drugs.

Functional assignment to hypothetical proteins in Orientia tsutsugamushistrain Ikeda

Orientia tsutsugamushi(O. tsutsugamushi) is an intracellular bacterial pathogen which causes zoonosis scrub typhus in humans. Genome of O. tsutsugamushi strain Ikeda contains 214 hypothetical proteins (HPs) which is nearly 20% of the total proteins. Domain and family based functional analysis of HPs results in the annotation of 44 hypothetical proteins. The annotated HPs were classified in to five main classes namely, gene expression and regulation, transport, metabolism, cell signaling and proteolysis. Thus, computational analysis of HPs helps to understand their putative roles in various biological and cellular processes, including pathogenesis for further consideration as potential therapeutic targets.

Machine and Deep Learning for Prediction of Subcellular Localization

Protein subcellular localization prediction (PSLP), which plays an important role in the field of computational biology, identifies the position and function of proteins in cells without expensive cost and laborious effort. In the past few decades, various methods with different algorithms have been proposed in solving the problem of subcellular localization prediction; machine learning and deep learning constitute a large portion among those proposed methods. In order to provide an overview about those methods, the first part of this article will be a brief review of several state-of-the-art machine learning methods on subcellular localization prediction; then the materials used by subcellular localization prediction is described and a simple prediction method, that takes protein sequences as input and utilizes a convolutional neural network as the classifier, is introduced. At last, a list of notes is provided to indicate the major problems that may occur with this method.

Surface proteomics and label-free quantification of Leptospira interrogans serovar Pomona

Leptospirosis is a re-emerging zoonosis with a global distribution. Surface-exposed outer membrane proteins (SE-OMPs) are crucial for bacterial-host interactions. SE-OMPs locate and expose their epitope on cell surface where is easily accessed by host molecules. This study aimed to screen for surface-exposed proteins and their abundance profile of pathogenic Leptospira interrogans serovar Pomona. Two complementary approaches, surface biotinylation and surface proteolytic shaving, followed by liquid chromatography tandem-mass spectrometry (LC-MS/MS) were employed to identify SE-OMPs of intact leptospires. For quantitative comparison, in-depth label-free analysis of SE-OMPs obtained from each method was performed using MaxQuant. The total number of proteins identified was 1,001 and 238 for surface biotinylation and proteinase K shaving, respectively. Among these, 39 were previously known SE-OMPs and 68 were predicted to be localized on the leptospiral surface. Based on MaxQuant analysis for relative quantification, six known SE-OMPs including EF- Tu, LipL21, LipL41, LipL46, Loa22, and OmpL36, and one predicted SE-OMPs, LipL71 were found in the 20 most abundant proteins, in which LipL41 was the highest abundant SE-OMP. Moreover, uncharacterized LIC14011 protein (LIP3228 ortholog in serovar Pomona) was identified as a novel predicted surface βb-OMP. High-abundance leptospiral SE-OMPs identified in this study may play roles in virulence and infection and are potential targets for development of vaccine or diagnostic tests for leptospirosis.

Predicting subcellular location of protein with evolution information and sequence-based deep learning

BACKGROUND

Protein subcellular localization prediction plays an important role in biology research. Since traditional methods are laborious and time-consuming, many machine learning-based prediction methods have been proposed. However, most of the proposed methods ignore the evolution information of proteins. In order to improve the prediction accuracy, we present a deep learning-based method to predict protein subcellular locations.

RESULTS

Our method utilizes not only amino acid compositions sequence but also evolution matrices of proteins. Our method uses a bidirectional long short-term memory network that processes the entire protein sequence and a convolutional neural network that extracts features from protein sequences. The position specific scoring matrix is used as a supplement to protein sequences. Our method was trained and tested on two benchmark datasets. The experiment results show that our method yields accurate results on the two datasets with an average precision of 0.7901, ranking loss of 0.0758 and coverage of 1.2848.

CONCLUSION

The experiment results show that our method outperforms five methods currently available. According to those experiments, we can see that our method is an acceptable alternative to predict protein subcellular location.

Pan-proteome profiling of emerging and re-emerging zoonotic pathogen Orientia tsutsugamushi for getting insight into microbial pathogenesis

With the occurrence and evolution of antibiotic and multidrug resistance in bacteria most of the existing remedies are becoming ineffective. The pan-proteome exploration of the bacterial pathogens helps to identify the wide spectrum therapeutic targets which will be effective against all strains in a species. The current study is focused on the pan-proteome profiling of zoonotic pathogen Orientia tsutsugamushi (Ott) for the identification of potential therapeutic targets. The pan-proteome of Ott is estimated to be extensive in nature that has 1429 protein clusters, out of which 694 were core, 391 were accessory, and 344 were unique. It was revealed that 622 proteins were essential, 222 proteins were virulent factors, and 42 proteins were involved in antibiotic resistance. The potential therapeutic targets were further classified into eleven broad classes among which gene expression and regulation, transport, and metabolism were dominant. The biological interactome analysis of therapeutic targets revealed that an ample amount of interactions were present among the proteins involved in DNA replication, ribosome assembly, cellwall metabolism, cell division, and antimicrobial resistance. The predicted therapeutic targets from the pan-proteome of Ott are involved in various biological processes, virulence, and antibiotic resistance; hence envisioned as potential candidates for drug discovery to combat scrub typhus.

PSORTdb 4.0: expanded and redesigned bacterial and archaeal protein subcellular localization database incorporating new secondary localizations

Protein subcellular localization (SCL) is important for understanding protein function, genome annotation, and aids identification of potential cell surface diagnostic markers, drug targets, or vaccine components. PSORTdb comprises ePSORTdb, a manually curated database of experimentally verified protein SCLs, and cPSORTdb, a pre-computed database of PSORTb-predicted SCLs for NCBI's RefSeq deduced bacterial and archaeal proteomes. We now report PSORTdb 4.0 (http://db.psort.org/). It features a website refresh, in particular a more user-friendly database search. It also addresses the need to uniquely identify proteins from NCBI genomes now that GI numbers have been retired. It further expands both ePSORTdb and cPSORTdb, including additional data about novel secondary localizations, such as proteins found in bacterial outer membrane vesicles. Protein predictions in cPSORTdb have increased along with the number of available microbial genomes, from approximately 13 million when PSORTdb 3.0 was released, to over 66 million currently. Now, analyses of both complete and draft genomes are included. This expanded database will be of wide use to researchers developing SCL predictors or studying diverse microbes, including medically, agriculturally and industrially important species that have both classic or atypical cell envelope structures or vesicles.

Bioinformatic Exploration of Metal-Binding Proteome of Zoonotic Pathogen Orientia tsutsugamushi

Metal ions are involved in many essential biological processes and are crucial for the survival of all organisms. Identification of metal-binding proteins (MBPs) of human affecting pathogens may provide the blueprint for understanding biological metal usage and their putative roles in pathogenesis. This study is focused on the analysis of MBPs from Orientia tsutsugamushi (Ott), a causal agent of scrub typhus in humans. A total of 321 proteins were predicted as putative MBPs, based on sequence search and three-dimensional structure analysis. Majority of proteins could bind with magnesium, and the order of metal binding was Mg > Ca > Zn > Mn > Fe > Cd > Ni > Co > Cu, respectively. The predicted MBPs were functionally classified into nine broad classes. Among them, gene expression and regulation, metabolism, cell signaling, and transport classes were dominant. It was noted that the putative MBPs were localized in all subcellular compartments of Ott, but majorly found in the cytoplasm. Additionally, it was revealed that out of 321 predicted MBPs 245 proteins were putative bacterial toxins and among them, 98 proteins were nonhomologous to human proteome. Sixty putative MBPs showed the ability to interact with drug or drug-like molecules, which indicate that they may be used as broad-spectrum drug targets. These predicted MBPs from Ott could play vital role(s) in various cellular activities and virulence, hence may serve as plausible therapeutic targets to design metal-based drugs to curtail its infection.

Burkholderia thailandensis strain E555 is a surrogate for the investigation of Burkholderia pseudomallei replication and survival in macrophages

Background

Burkholderia pseudomallei is a human pathogen causing severe infections in tropical and subtropical regions and is classified as a bio-threat agent. B. thailandensis strain E264 has been proposed as less pathogenic surrogate for understanding the interactions of B. pseudomallei with host cells.

Results

We show that, unlike B. thailandensis strain E264, the pattern of growth of B. thailandensis strain E555 in macrophages is similar to that of B. pseudomallei. We have genome sequenced B. thailandensis strain E555 and using the annotated sequence identified genes and proteins up-regulated during infection. Changes in gene expression identified more of the known B. pseudomallei virulence factors than changes in protein levels and used together we identified 16% of the currently known B. pseudomallei virulence factors. These findings demonstrate the utility of B. thailandensis strain E555 to study virulence of B. pseudomallei.

Conclusions

A weakness of studies using B. thailandensis as a surrogate for B. pseudomallei is that the strains used replicate at a slower rate in infected cells. We show that the pattern of growth of B. thailandensis strain E555 in macrophages closely mirrors that of B. pseudomallei. Using this infection model we have shown that virulence factors of B. pseudomallei can be identified as genes or proteins whose expression is elevated on the infection of macrophages. This finding confirms the utility of B. thailandensis strain E555 as a surrogate for B. pseudomallei and this strain should be used for future studies on virulence mechanisms.

Electronic supplementary material

The online version of this article (10.1186/s12866-019-1469-8) contains supplementary material, which is available to authorized users.

Comparative proteome analysis reveals pathogen specific outer membrane proteins of Leptospira

Proteomes of pathogenic Leptospira interrogans and L. borgpetersenii and the saprophytic L. biflexa were filtered through computational tools to identify Outer Membrane Proteins (OMPs) that satisfy the required biophysical parameters for their presence on the outer membrane. A total of 133, 130, and 144 OMPs were identified in L. interrogans, L. borgpetersenii, and L. biflexa, respectively, which forms approximately 4% of proteomes. A holistic analysis of transporting and pathogenic characteristics of OMPs together with Clusters of Orthologous Groups (COGs) among the OMPs and their distribution across 3 species was made and put forward a set of 21 candidate OMPs specific to pathogenic leptospires. It is also found that proteins homologous to the candidate OMPs were also present in other pathogenic species of leptospires. Six OMPs from L. interrogans and 2 from L. borgpetersenii observed to have similar COGs while those were not found in any intermediate or saprophytic forms. These OMPs appears to have role in infection and pathogenesis and useful for anti-leptospiral strategies.

Protein Sorting Prediction

Many computational methods are available for predicting protein sorting in bacteria. When comparing them, it is important to know that they can be grouped into three fundamentally different approaches: signal-based, global-property-based and homology-based prediction. In this chapter, the strengths and drawbacks of each of these approaches is described through many examples of methods that predict secretion, integration into membranes, or subcellular locations in general. The aim of this chapter is to provide a user-level introduction to the field with a minimum of computational theory.

Predicting Subcellular Localization of Proteins by Bioinformatic Algorithms

When predicting the subcellular localization of proteins from their amino acid sequences, there are basically three approaches: signal-based, global property-based, and homology-based. Each of these has its advantages and drawbacks, and it is important when comparing methods to know which approach was used. Various statistical and machine learning algorithms are used with all three approaches, and various measures and standards are employed when reporting the performances of the developed methods. This chapter presents a number of available methods for prediction of sorting signals and subcellular localization, but rather than providing a checklist of which predictors to use, it aims to function as a guide for critical assessment of prediction methods.

PSORTdb: expanding the bacteria and archaea protein subcellular localization database to better reflect diversity in cell envelope structures

Protein subcellular localization (SCL) is important for understanding protein function, genome annotation, and has practical applications such as identification of potential vaccine components or diagnostic/drug targets. PSORTdb (http://db.psort.org) comprises manually curated SCLs for proteins which have been experimentally verified (ePSORTdb), as well as pre-computed SCL predictions for deduced proteomes from bacterial and archaeal complete genomes available from NCBI (cPSORTdb). We now report PSORTdb 3.0. It features improvements increasing user-friendliness, and further expands both ePSORTdb and cPSORTdb with a focus on improving protein SCL data in cases where it is most difficult—proteins associated with non-classical Gram-positive/Gram-negative/Gram-variable cell envelopes. ePSORTdb data curation was expanded, including adding in additional cell envelope localizations, and incorporating markers for cPSORTdb to automatically computationally identify if new genomes to be analysed fall into certain atypical cell envelope categories (i.e. Deinococcus-Thermus, Thermotogae, Corynebacteriales/Corynebacterineae, including Mycobacteria). The number of predicted proteins in cPSORTdb has increased from 3 700 000 when PSORTdb 2.0 was released to over 13 000 000 currently. PSORTdb 3.0 will be of wider use to researchers studying a greater diversity of monoderm or diderm microbes, including medically, agriculturally and industrially important species that have non-classical outer membranes or other cell envelope features.

Identification of Outer Membrane and Exoproteins of Carbapenem-Resistant Multilocus Sequence Type 258 Klebsiella pneumoniae

Carbapenem-resistant Klebsiella pneumoniae strains have emerged as a cause of life-threatening infections in susceptible individuals (e.g., transplant recipients and critically ill patients). Strains classified as multilocus sequence type (ST) 258 are among the most prominent causes of carbapenem-resistant K. pneumoniae infections worldwide, but the basis for the success of this lineage remains incompletely determined. To gain a more comprehensive view of the molecules potentially involved in the success of ST258, we used a proteomics approach to identify surface-associated and culture supernatant proteins produced by ST258. Protein samples were prepared from varied culture conditions in vitro, and were analyzed by a combination of two-dimensional electrophoresis and liquid chromatography followed by tandem mass spectrometry (LC-MS/MS). We identified a total of 193 proteins in outer membrane preparations from bacteria cultured in Luria-Bertani broth (LB) or RPMI 1640 tissue culture media (RPMI). Compared with LB, several iron-acquisition proteins, including IutA, HmuR, HmuS, CirA, FepA, FitA, FoxA, FhuD, and YfeX, were more highly expressed in RPMI. Of the 177 proteins identified in spent media, only the fimbrial subunit, MrkA, was predicted to be extracellular, a finding that suggests few proteins (or a limited quantity) are freely secreted by ST258. Notably, we discovered 203 proteins not reported in previous K. pneumoniae proteome studies. In silico modeling of proteins with unknown function revealed several proteins with beta-barrel transmembrane structures typical of porins, as well as possible host-interacting proteins. Taken together, these findings contribute several new targets for the mechanistic study of drug-resistance and pathogenesis by ST258 K. pneumoniae isolates.

Identification of host-microbe interaction factors in the genomes of soft rot-associated pathogens Dickeya dadantii 3937 and Pectobacterium carotovorum WPP14 with supervised machine learning

Background

A wealth of genome sequences has provided thousands of genes of unknown function, but identification of functions for the large numbers of hypothetical genes in phytopathogens remains a challenge that impacts all research on plant-microbe interactions. Decades of research on the molecular basis of pathogenesis focused on a limited number of factors associated with long-known host-microbe interaction systems, providing limited direction into this challenge. Computational approaches to identify virulence genes often rely on two strategies: searching for sequence similarity to known host-microbe interaction factors from other organisms, and identifying islands of genes that discriminate between pathogens of one type and closely related non-pathogens or pathogens of a different type. The former is limited to known genes, excluding vast collections of genes of unknown function found in every genome. The latter lacks specificity, since many genes in genomic islands have little to do with host-interaction.

Result

In this study, we developed a supervised machine learning approach that was designed to recognize patterns from large and disparate data types, in order to identify candidate host-microbe interaction factors. The soft rot Enterobacteriaceae strains Dickeya dadantii 3937 and Pectobacterium carotovorum WPP14 were used for development of this tool, because these pathogens are important on multiple high value crops in agriculture worldwide and more genomic and functional data is available for the Enterobacteriaceae than any other microbial family. Our approach achieved greater than 90% precision and a recall rate over 80% in 10-fold cross validation tests.

Conclusion

Application of the learning scheme to the complete genome of these two organisms generated a list of roughly 200 candidates, many of which were previously not implicated in plant-microbe interaction and many of which are of completely unknown function. These lists provide new targets for experimental validation and further characterization, and our approach presents a promising pattern-learning scheme that can be generalized to create a resource to study host-microbe interactions in other bacterial phytopathogens.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2164-15-508) contains supplementary material, which is available to authorized users.

Protein (multi-)location prediction: using location inter-dependencies in a probabilistic framework

Motivation

Knowing the location of a protein within the cell is important for understanding its function, role in biological processes, and potential use as a drug target. Much progress has been made in developing computational methods that predict single locations for proteins. Most such methods are based on the over-simplifying assumption that proteins localize to a single location. However, it has been shown that proteins localize to multiple locations. While a few recent systems attempt to predict multiple locations of proteins, their performance leaves much room for improvement. Moreover, they typically treat locations as independent and do not attempt to utilize possible inter-dependencies among locations. Our hypothesis is that directly incorporating inter-dependencies among locations into both the classifier-learning and the prediction process can improve location prediction performance.

Results

We present a new method and a preliminary system we have developed that directly incorporates inter-dependencies among locations into the location-prediction process of multiply-localized proteins. Our method is based on a collection of Bayesian network classifiers, where each classifier is used to predict a single location. Learning the structure of each Bayesian network classifier takes into account inter-dependencies among locations, and the prediction process uses estimates involving multiple locations. We evaluate our system on a dataset of single- and multi-localized proteins (the most comprehensive protein multi-localization dataset currently available, derived from the DBMLoc dataset). Our results, obtained by incorporating inter-dependencies, are significantly higher than those obtained by classifiers that do not use inter-dependencies. The performance of our system on multi-localized proteins is comparable to a top performing system (YLoc⁺), without being restricted only to location-combinations present in the training set.

Genome characterization of a novel Burkholderia cepacia complex genomovar isolated from dieback affected mango orchards

We characterized the genome of the antibiotic resistant, caseinolytic and non-hemolytic Burkholderia sp. strain TJI49, isolated from mango trees (Mangifera indica L.) with dieback disease. This isolate produced severe disease symptoms on the indicator plants. Next generation DNA sequencing and short-read assembly generated the 60X deep 7,631,934 nucleotide draft genome of Burkholderia sp. TJI49 which comprised three chromosomes and at least one mega plasmid. Genome annotation studies revealed a total 8,992 genes, out of which 8,940 were protein coding genes. Comparative genomics and phylogenetics identified Burkholderia sp. TJI49 as a distinct species of Burkholderia cepacia complex (BCC), closely related to B. multivorans ATCC17616. Genome-wide sequence alignment of this isolate with replicons of BCC members showed conservation of core function genes but considerable variations in accessory genes. Subsystem-based gene annotation identified the active presence of wide spread colonization island and type VI secretion system in Burkholderia sp. TJI49. Sequence comparisons revealed (a) 28 novel ORFs that have no database matches and (b) 23 ORFs with orthologues in species other than Burkholderia, indicating horizontal gene transfer events. Fold recognition of novel ORFs identified genes encoding pertactin autotransporter-like proteins (a constituent of type V secretion system) and Hap adhesion-like proteins (involved in cell-cell adhesion) in the genome of Burkholderia sp. TJI49. The genomic characterization of this isolate provided additional information related to the 'pan-genome' of Burkholderia species.

Reverse vaccinology in the 21st century: improvements over the original design

Reverse vaccinology (RV), the first application of genomic technologies in vaccine research, represented a major revolution in the process of discovering novel vaccines. By determining their entire antigenic repertoire, researchers could identify protective targets and design efficacious vaccines for pathogens where conventional approaches had failed. Bexsero, the first vaccine developed using RV, has recently received positive opinion from the European Medicines Agency. The use of RV initiated a cascade of changes that affected the entire vaccine development process, shifting the focus from the identification of a list of vaccine candidates to the definition of a set of high throughput screens to reduce the need for costly and labor intensive tests in animal models. It is now clear that a deep understanding of the epidemiology of vaccine candidates, and their regulation and role in host-pathogen interactions, must become an integral component of the screening workflow. Far from being outdated by technological advancements, RV still represents a paradigm of how high-throughput technologies and scientific insight can be integrated into biotechnology research.

Combining machine learning and homology-based approaches to accurately predict subcellular localization in Arabidopsis

A complete map of the Arabidopsis (Arabidopsis thaliana) proteome is clearly a major goal for the plant research community in terms of determining the function and regulation of each encoded protein. Developing genome-wide prediction tools such as for localizing gene products at the subcellular level will substantially advance Arabidopsis gene annotation. To this end, we performed a comprehensive study in Arabidopsis and created an integrative support vector machine-based localization predictor called AtSubP (for Arabidopsis subcellular localization predictor) that is based on the combinatorial presence of diverse protein features, such as its amino acid composition, sequence-order effects, terminal information, Position-Specific Scoring Matrix, and similarity search-based Position-Specific Iterated-Basic Local Alignment Search Tool information. When used to predict seven subcellular compartments through a 5-fold cross-validation test, our hybrid-based best classifier achieved an overall sensitivity of 91% with high-confidence precision and Matthews correlation coefficient values of 90.9% and 0.89, respectively. Benchmarking AtSubP on two independent data sets, one from Swiss-Prot and another containing green fluorescent protein- and mass spectrometry-determined proteins, showed a significant improvement in the prediction accuracy of species-specific AtSubP over some widely used "general" tools such as TargetP, LOCtree, PA-SUB, MultiLoc, WoLF PSORT, Plant-PLoc, and our newly created All-Plant method. Cross-comparison of AtSubP on six nontrained eukaryotic organisms (rice [Oryza sativa], soybean [Glycine max], human [Homo sapiens], yeast [Saccharomyces cerevisiae], fruit fly [Drosophila melanogaster], and worm [Caenorhabditis elegans]) revealed inferior predictions. AtSubP significantly outperformed all the prediction tools being currently used for Arabidopsis proteome annotation and, therefore, may serve as a better complement for the plant research community. A supplemental Web site that hosts all the training/testing data sets and whole proteome predictions is available at http://bioinfo3.noble.org/AtSubP/.

PSORTb 3.0: improved protein subcellular localization prediction with refined localization subcategories and predictive capabilities for all prokaryotes

Motivation: PSORTb has remained the most precise bacterial protein subcellular localization (SCL) predictor since it was first made available in 2003. However, the recall needs to be improved and no accurate SCL predictors yet make predictions for archaea, nor differentiate important localization subcategories, such as proteins targeted to a host cell or bacterial hyperstructures/organelles. Such improvements should preferably be encompassed in a freely available web-based predictor that can also be used as a standalone program.

Results: We developed PSORTb version 3.0 with improved recall, higher proteome-scale prediction coverage, and new refined localization subcategories. It is the first SCL predictor specifically geared for all prokaryotes, including archaea and bacteria with atypical membrane/cell wall topologies. It features an improved standalone program, with a new batch results delivery system complementing its web interface. We evaluated the most accurate SCL predictors using 5-fold cross validation plus we performed an independent proteomics analysis, showing that PSORTb 3.0 is the most accurate but can benefit from being complemented by Proteome Analyst predictions.

Availability:http://www.psort.org/psortb (download open source software or use the web interface).

Contact:psort-mail@sfu.ca

Supplementary Information:Supplementary data are availableat Bioinformatics online.

Bioinformatics predictions of localization and targeting

One of the major challenges in the post-genomic era with hundreds of genomes sequenced is the annotation of protein structure and function. Computational predictions of subcellular localization are an important step toward this end. The development of computational tools that predict targeting and localization has, therefore, been a very active area of research, in particular since the first release of the groundbreaking program PSORT in 1991. The most reliable means of annotating protein structure and function remains homology-based inference, i.e. the transfer of experimental annotations from one protein to its homologs. However, annotations about localization demonstrate how much can be gained from advanced machine learning: more proteins can be annotated more reliably. Contemporary computational tools for the annotation of protein targeting include automatic methods that mine the textual information from the biological literature and molecular biology databases. Some machine learning-based methods that accurately predict features of sorting signals and that use sequence-derived features to predict localization have reached remarkable levels of performance. Sustained prediction accuracy has increased by more than 30 percentage points over the last decade. Here, we review some of the most recent methods for the prediction of subcellular localization and protein targeting that contributed toward this breakthrough.

Overcoming function annotation errors in the Gram-positive pathogen Streptococcus suis by a proteomics-driven approach

Background

Annotation of protein-coding genes is a key step in sequencing projects. Protein functions are mainly assigned on the basis of the amino acid sequence alone by searching of homologous proteins. However, fully automated annotation processes often lead to wrong prediction of protein functions, and therefore time-intensive manual curation is often essential. Here we describe a fast and reliable way to correct function annotation in sequencing projects, focusing on surface proteomes. We use a proteomics approach, previously proven to be very powerful for identifying new vaccine candidates against Gram-positive pathogens. It consists of shaving the surface of intact cells with two proteases, the specific cleavage-site trypsin and the unspecific proteinase K, followed by LC/MS/MS analysis of the resulting peptides. The identified proteins are contrasted by computational analysis and their sequences are inspected to correct possible errors in function prediction.

Results

When applied to the zoonotic pathogen Streptococcus suis, of which two strains have been recently sequenced and annotated, we identified a set of surface proteins without cytoplasmic contamination: all the proteins identified had exporting or retention signals towards the outside and/or the cell surface, and viability of protease-treated cells was not affected. The combination of both experimental evidences and computational methods allowed us to determine that two of these proteins are putative extracellular new adhesins that had been previously attributed a wrong cytoplasmic function. One of them is a putative component of the pilus of this bacterium.

Conclusion

We illustrate the complementary nature of laboratory-based and computational methods to examine in concert the localization of a set of proteins in the cell, and demonstrate the utility of this proteomics-based strategy to experimentally correct function annotation errors in sequencing projects. This approach also contributes to provide strong experimental evidences that can be used to annotate those proteins for which a Gene Ontology (GO) term has not been assigned so far. Function annotation correction would then improve the identification of surface-associated proteins in bacterial pathogens, thus accelerating the discovery of new vaccines in infectious disease research.

Computational and experimental approaches to chart the Escherichia coli cell-envelope-associated proteome and interactome

The bacterial cell-envelope consists of a complex arrangement of lipids, proteins and carbohydrates that serves as the interface between a microorganism and its environment or, with pathogens, a human host. Escherichia coli has long been investigated as a leading model system to elucidate the fundamental mechanisms underlying microbial cell-envelope biology. This includes extensive descriptions of the molecular identities, biochemical activities and evolutionary trajectories of integral transmembrane proteins, many of which play critical roles in infectious disease and antibiotic resistance. Strikingly, however, only half of the c. 1200 putative cell-envelope-related proteins of E. coli currently have experimentally attributed functions, indicating an opportunity for discovery. In this review, we summarize the state of the art of computational and proteomic approaches for determining the components of the E. coli cell-envelope proteome, as well as exploring the physical and functional interactions that underlie its biogenesis and functionality. We also provide a comprehensive comparative benchmarking analysis on the performance of different bioinformatic and proteomic methods commonly used to determine the subcellular localization of bacterial proteins.

Penalized feature selection and classification in bioinformatics

In bioinformatics studies, supervised classification with high-dimensional input variables is frequently encountered. Examples routinely arise in genomic, epigenetic and proteomic studies. Feature selection can be employed along with classifier construction to avoid over-fitting, to generate more reliable classifier and to provide more insights into the underlying causal relationships. In this article, we provide a review of several recently developed penalized feature selection and classification techniques--which belong to the family of embedded feature selection methods--for bioinformatics studies with high-dimensional input. Classification objective functions, penalty functions and computational algorithms are discussed. Our goal is to make interested researchers aware of these feature selection and classification methods that are applicable to high-dimensional bioinformatics data.

Genome-wide subcellular localization of putative outer membrane and extracellular proteins in Leptospira interrogans serovar Lai genome using bioinformatics approaches

BACKGROUND

In bacterial pathogens, both cell surface-exposed outer membrane proteins and proteins secreted into the extracellular environment play crucial roles in host-pathogen interaction and pathogenesis. Considerable efforts have been made to identify outer membrane (OM) and extracellular (EX) proteins produced by Leptospira interrogans, which may be used as novel targets for the development of infection markers and leptospirosis vaccines.

RESULT

In this study we used a novel computational framework based on combined prediction methods with deduction concept to identify putative OM and EX proteins encoded by the Leptospira interrogans genome. The framework consists of the following steps: (1) identifying proteins homologous to known proteins in subcellular localization databases derived from the "consensus vote" of computational predictions, (2) incorporating homology based search and structural information to enhance gene annotation and functional identification to infer the specific structural characters and localizations, and (3) developing a specific classifier for cytoplasmic proteins (CP) and cytoplasmic membrane proteins (CM) using Linear discriminant analysis (LDA). We have identified 114 putative EX and 63 putative OM proteins, of which 41% are conserved or hypothetical proteins containing sequence and/or protein folding structures similar to those of known EX and OM proteins.

CONCLUSION

Overall results derived from the combined computational analysis correlate with the available experimental evidence. This is the most extensive in silico protein subcellular localization identification to date for Leptospira interrogans serovar Lai genome that may be useful in protein annotation, discovery of novel genes and understanding the biology of Leptospira.

LocateP: genome-scale subcellular-location predictor for bacterial proteins

Background

In the past decades, various protein subcellular-location (SCL) predictors have been developed. Most of these predictors, like TMHMM 2.0, SignalP 3.0, PrediSi and Phobius, aim at the identification of one or a few SCLs, whereas others such as CELLO and Psortb.v.2.0 aim at a broader classification. Although these tools and pipelines can achieve a high precision in the accurate prediction of signal peptides and transmembrane helices, they have a much lower accuracy when other sequence characteristics are concerned. For instance, it proved notoriously difficult to identify the fate of proteins carrying a putative type I signal peptidase (SPIase) cleavage site, as many of those proteins are retained in the cell membrane as N-terminally anchored membrane proteins. Moreover, most of the SCL classifiers are based on the classification of the Swiss-Prot database and consequently inherited the inconsistency of that SCL classification. As accurate and detailed SCL prediction on a genome scale is highly desired by experimental researchers, we decided to construct a new SCL prediction pipeline: LocateP.

Results

LocateP combines many of the existing high-precision SCL identifiers with our own newly developed identifiers for specific SCLs. The LocateP pipeline was designed such that it mimics protein targeting and secretion processes. It distinguishes 7 different SCLs within Gram-positive bacteria: intracellular, multi-transmembrane, N-terminally membrane anchored, C-terminally membrane anchored, lipid-anchored, LPxTG-type cell-wall anchored, and secreted/released proteins. Moreover, it distinguishes pathways for Sec- or Tat-dependent secretion and alternative secretion of bacteriocin-like proteins. The pipeline was tested on data sets extracted from literature, including experimental proteomics studies. The tests showed that LocateP performs as well as, or even slightly better than other SCL predictors for some locations and outperforms current tools especially where the N-terminally anchored and the SPIase-cleaved secreted proteins are concerned. Overall, the accuracy of LocateP was always higher than 90%. LocateP was then used to predict the SCLs of all proteins encoded by completed Gram-positive bacterial genomes. The results are stored in the database LocateP-DB [1].

Conclusion

LocateP is by far the most accurate and detailed protein SCL predictor for Gram-positive bacteria currently available.

Computer-aided biotechnology: from immuno-informatics to reverse vaccinology

Genome sequences from many organisms, including humans, have been completed, and high-throughput analyses have produced burgeoning volumes of 'omics' data. Bioinformatics is crucial for the management and analysis of such data and is increasingly used to accelerate progress in a wide variety of large-scale and object-specific functional analyses. Refined algorithms enable biotechnologists to follow 'computer-aided strategies' based on experiments driven by high-confidence predictions. In order to address compound problems, current efforts in immuno-informatics and reverse vaccinology are aimed at developing and tuning integrative approaches and user-friendly, automated bioinformatics environments. This will herald a move to 'computer-aided biotechnology': smart projects in which time-consuming and expensive large-scale experimental approaches are progressively replaced by prediction-driven investigations.

Computational prediction of subcellular localization

It is widely recognized that much of the information for determining the final subcellular localization of proteins is found in their amino acid sequences. Thus the prediction of protein localization sites is of both theoretical and practical interest. In most cases, the prediction has been attempted in two ways: one is based on the knowledge of experimentally characterized targeting signals, while the other utilizes the statistical differences of general sequence characteristics, such as amino acid composition, between localization sites. Both approaches have limitations, and it is recommended to check the results of various prediction methods based on different principles as well as training data. Recently, increased proteomic analyses of localization sites have provided new data to assess the current status of predictive methods. In this chapter we discuss these issues and close with an example illustrating the use of the WoLF PSORT web server for localization prediction.

In silico and microarray-based genomic approaches to identifying potential vaccine candidates against Leptospira interrogans

BACKGROUND

Currently available vaccines against leptospirosis are of low efficacy, have an unacceptable side-effect profile, do not induce long-term protection, and provide no cross-protection against the different serovars of pathogenic leptospira. The current major focus in leptospirosis research is to discover conserved protective antigens that may elicit longer-term protection against a broad range of Leptospira. There is a need to screen vaccine candidate genes in the genome of Leptospira interrogans.

RESULTS

Bioinformatics, comparative genomic hybridization (CGH) analysis and transcriptional analysis were used to identify vaccine candidates in the genome of L. interrogans serovar Lai strain #56601. Of a total of 4727 open reading frames (ORFs), 616 genes were predicted to encode surface-exposed proteins by P-CLASSIFIER combined with signal peptide prediction, alpha-helix transmembrane topology prediction, integral beta-barrel outer membrane protein and lipoprotein prediction, as well as by retaining the genes shared by the two sequenced L. interrogans genomes and by subtracting genes with human homologues. A DNA microarray of L. interrogans strain #56601 was constructed for CGH analysis and transcriptome analysis in vitro. Three hundred and seven differential genes were identified in ten pathogenic serovars by CGH; 1427 genes had high transcriptional levels (Cy3 signal > or = 342 and Cy5 signal > or = 363.5, respectively). There were 565 genes in the intersection between the set encoding surface-exposed proteins and the set of 307 differential genes. The number of genes in the intersection between this set of 565 and the set of 1427 highly transcriptionally active genes was 226. These 226 genes were thus identified as putative vaccine candidates. The proteins encoded by these genes are not only potentially surface-exposed in the bacterium, but also conserved in two sequenced L. interrogans. Moreover, these genes are conserved among ten epidemic serovars in China and have high transcriptional levels in vitro.

CONCLUSION

Of the 4727 ORFs in the genome of L. interrogans, 226 genes were identified as vaccine candidates by bioinformatics, CGH and transcriptional analysis on the basis of the theory of reverse vaccinology. The proteins encoded by these genes might be useful as vaccine candidates as well as for diagnosis of leptospirosis.

Methods for predicting bacterial protein subcellular localization

The computational prediction of the subcellular localization of bacterial proteins is an important step in genome annotation and in the search for novel vaccine or drug targets. Since the 1991 release of PSORT I--the first comprehensive algorithm to predict bacterial protein localization--many other localization prediction tools have been developed. These methods offer significant improvements in predictive performance over PSORT I and the accuracy of some methods now rivals that of certain high-throughput laboratory methods for protein localization identification.