PathogenFinder--distinguishing friend from foe using bacterial whole genome sequence data.
Cosentino S, Voldby Larsen M, Møller Aarestrup F, Lund O. · 2013
Read PathogenFinder--distinguishing friend from foe using bacterial whole genome sequence data. by Cosentino S, Voldby Larsen M, Møller Aarestrup F, Lund O. (2013) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc3810466.txt.
Introduction
Cosentino and colleagues introduced in PLoS ONE in 2013 PathogenFinder, a web tool that predicts from a bacterium's genome whether it is likely to harm people. The tool accepts a proteome, a genome, or raw sequencing reads and works across all bacterial groups.
Instead of checking for known disease genes, the method scores protein families drawn from 885 complete genomes, 372 from human pathogens and 513 from harmless strains. On an independent test set it classified 398 of 449 bacteria correctly, an accuracy of 88.6 percent. Because the families were picked without regard to known function, the authors say the tool may also point to disease factors nobody has named yet.
Catalog record
| CREATOR | Cosentino S, Voldby Larsen M, Møller Aarestrup F, Lund O. |
|---|---|
| YEAR | 2013 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
PathogenFinder - Distinguishing Friend from Foe Using Bacterial Whole Genome Sequence Data By Salvatore Cosentino, Mette Voldby Larsen, Frank Møller Aarestrup, Ole Lund, Jason D. Barbour Published in: PLoS ONE DOI: 10.1371/journal.pone.0077302 ABSTRACT Although the majority of bacteria are harmless or even beneficial to their host, others are highly virulent and can cause serious diseases, and even death. Due to the constantly decreasing cost of high-throughput sequencing there are now many completely sequenced genomes available from both human pathogenic and innocuous strains. The data can be used to identify gene families that correlate with pathogenicity and to develop tools to predict the pathogenicity of newly sequenced strains, investigations that previously were mainly done by means of more expensive and time consuming experimental approaches. We describe PathogenFinder (http://cge.cbs.dtu.dk/services/PathogenFinder/), a web-server for the prediction of bacterial pathogenicity by analysing the input proteome, genome, or raw reads provided by the user. The method relies on groups of proteins, created without regard to their annotated function or known involvement in pathogenicity. The method has been built to work with all taxonomic groups of bacteria and using the entire training-set, achieved an accuracy of 88.6% on an independent test-set, by correctly classifying 398 out of 449 completely sequenced bacteria. The approach here proposed is not biased on sets of genes known to be associated with pathogenicity, thus the approach could aid the discovery of novel pathogenicity factors. Furthermore the pathogenicity prediction web-server could be used to isolate the potential pathogenic features of both known and unknown strains. INTRODUCTION Every year more than 15 millions deaths are the direct cause of infectious diseases, many of which are due to bacterial infections. Each year an estimated 1.3 million people die of tuberculosis and 0.2 millions of pertussis, while diarrhoea accounts for more than 2.5 millions deaths, and is one of the leading causes of death in worldwide [1]. But not all bacteria are dangerous and many of them are innocuous or even beneficial to human. The gut of a healthy adult human contains thousand of different microbial species, many of which are beneficial to their host, providing functions for nutrition and development, and regulating the immune response [2], [3]. Nevertheless some bacterial species, like Escherichia coli, also include extremely deadly strains, causing for example diarrhoea, urinary tract infections, septicaemia etc. Thus identifying pathogenic strains and understanding the biological mechanisms that cause them to become pathogenic is highly important to perform timely interventions and design control strategies, including interventions such as restrictions on contaminated food products, isolation of patients, correct treatment and development of targeted vaccines. Ever since the 1880s the pathogenicity of bacteria have been assessed using Koch’s postulates, for human pathogens using animal models. During the last 2 decades many discoveries have shown that Koch’s postulates are not enough to decide if a given bacteria is pathogenic or not. The existence of diseases caused by bacteria that cannot grow in pure culture medium [4], [5], the discovery of polymicrobial diseases [6], the role of metagenomic microbiota in chronic diseases [7], and last but not least, the discovery of Horizontal Gene Transfer (HGT) responsible for the swapping of genetic material between bacteria [8] (regardless the pathogenicity), are all cases in which the postulates have short-comings. Already during his work with Vibrio cholerae Robert Koch himself discovered the shortcomings of animal models for correctly identifying human-specific pathogens. Thus, the use of animal models is not always reliable in defining if a given bacteria is human pathogenic. Moreover, assessing the pathogenicity by means of animal models or epidemiological studies is both time-consuming and expensive. Among the molecular features that a bacterium needs to infect and survive inside its host [9] are exotoxins, endotoxins, two components systems [10], adherence factors, secretion systems (I to IV type) [11], through which bacteria can inject their toxins into its hosts cells [12]. Plasmids, secretion systems, and antibiotic resistance genes are commonly present in both commensal and pathogenic strains, while toxins are usually only present in pathogenic strains. There are many databases available containing genes encoding toxins and virulence factors along with other genes traditionally associated with pathogenicity [13], [14]. One of the ways to classify a bacterium as human pathogenic using bioinformatics was (and still sometimes is) to look for some of these features in the genome of the isolate under investigation. Unluckily this approach is not always reliable, partly because of HGT, which causes these features to be exchanged among pathogenic and innocuous strains of the same [15] [16] or different species, an exchange which has been proved by the high amount of these features found in genomic islands [17]. Aside from the features directly associated to pathogenicity, there are also virulence “lifestyle” genes, important for the bacteria to survive inside the host and evade its immune system response [18] [19], and genes that are, for example, needed to activate other genes, which are important in the processes of pathogenesis, even though they do not directly determine virulence. All the issues related to the prediction of bacterial pathogenicity based on phylogeny has caused researchers to look for different solutions. The development of whole genome sequencing may open novel ways of predicting pathogenicity in bacterial species. In 1995 the genomes of Mycoplasma genitalium and Haemophilus influenzae [20], [21] were completely sequenced, and scientists started considering the possibility of studying the pathogenesis of bacteria based on their genome sequences [22]. This was the start of a revolution that has been continuing during the last decade with the advent of Second-Generation or Next-Generation Sequencing (NGS), leading to a continuous decrease in sequencing costs and a fast development of sequencing technologies. At present, many different high-throughput sequencing systems are available [23]–[25] and the number of completely sequenced bacteria amount to almost 2,400 including more than 1,800 that have been submitted to the International Nucleotide Sequence Database Collaboration (INSDC) (www.genomesonline.org, May 2013). A few methods have been proposed which make use of Support Vector Machines (SVM), BLAST or other bioinformatics tools to search for pathogenic features [26], [27] or predict bacterial pathogenicity [28] by searching in pre-computed databases of genes associated with pathogens. One shared aspect among these methods is the fact that they restrict their search to well known pathogenic features, missing out on the information that may be contained in the many genes with unknown function. Furthermore, the methods ignore genes that could be shared and specific among non-pathogenic organisms. When bacteria become pathogenic through HGT their lifestyle change and some of the genes may be inactivated or even lost to adapt to the new lifestyle [29], [30]. These genes are still present in non-pathogenic bacteria and hence could be used, together with the genes associated to pathogenicity, to separate dangerous bacteria from harmless ones. As an alternative to the above mentioned prediction methods, we here developed a novel approach, building on a previous study [31]. In this study we selected groups of genes which are frequently found either in human pathogenic bacteria or in the innocuous ones, and show that this is more effective than using global similarity. Since we did not make any pre-assumption on the genes contained in our training-sets, we are able to identify new proteins associated to pathogenicity and also