Performance assessment of genomic island prediction tools with an improved version of Design-Island

Genomic Islands (GIs) play an important role in the evolution and adaptation of prokaryotes. The origin and extent of ecological diversity of prokaryotes can be analyzed by comparing GIs across closely or distantly related prokaryotes. Understanding the importance of GI and to study the bacterial evolution, several GI prediction tools have been generated. An unsupervised method, Design-Island, was developed to identify GIs using Monte-Carlo statistical test on randomly selected segments of a chromosome. Here, in the present study Design-Island was modified with the incorporation of majority voting, multiple hypothesis testing correction. The performance of the modified version, Design-Island-II was tested and compared with the existing GI prediction tools. The performance assessment and benchmarking of the GI prediction tools require experimentally validated dataset, which is lacking. So, different datasets, generated or taken from literature were utilized to compare the sensitivity (SN), specificity (SP), precision (PPV) and accuracy (AC) of Design-Island-II. It showed substantial enhancement in term of SN, SP, PPV and AC, and significantly reduced the computation time of the algorithm. The performance of Design-Island-II has also been compared with several GI prediction tools using curated dataset of putative horizontally transferred genes. Design-Island-II showed the highest sensitivity and F1 score, comparable specificity, precision and accuracy in comparison to the other available methods. IslandViewer4 and Islander outperformed all the available methods in terms of AC and PPV respectively. Our study suggested Design-Island-II, IslandViewer4 and GIHunter among the top performing GI prediction tools considering both sensitivity and specificity of the methods.

SSG-LUGIA: Single Sequence based Genome Level Unsupervised Genomic Island Prediction Algorithm

BACKGROUND

Genomic Islands (GIs) are clusters of genes that are mobilized through horizontal gene transfer. GIs play a pivotal role in bacterial evolution as a mechanism of diversification and adaptation to different niches. Therefore, identification and characterization of GIs in bacterial genomes is important for understanding bacterial evolution. However, quantifying GIs is inherently difficult, and the existing methods suffer from low prediction accuracy and precision-recall trade-off. Moreover, several of them are supervised in nature, and thus, their applications to newly sequenced genomes are riddled with their dependency on the functional annotation of existing genomes.

RESULTS

We present SSG-LUGIA, a completely automated and unsupervised approach for identifying GIs and horizontally transferred genes. SSG-LUGIA is a novel method based on unsupervised anomaly detection technique, accompanied by further refinement using cues from signal processing literature. SSG-LUGIA leverages the atypical compositional biases of the alien genes to localize GIs in prokaryotic genomes. SSG-LUGIA was assessed on a large benchmark dataset `IslandPick' and on a set of 15 well-studied genomes in the literature and followed by a thorough analysis on the well-understood Salmonella typhi CT18 genome. Furthermore, the efficacy of SSG-LUGIA in identifying horizontally transferred genes was evaluated on two additional bacterial genomes, namely, those of Corynebacterium diphtheria NCTC13129 and Pseudomonas aeruginosa LESB58. SSG-LUGIA was examined on draft genomes and was demonstrated to be efficient as an ensemble method.

CONCLUSIONS

Our results indicate that SSG-LUGIA achieved superior performance in comparison to frequently used existing methods. Importantly, it yielded a better trade-off between precision and recall than the existing methods. Its nondependency on the functional annotation of genomes makes it suitable for analyzing newly sequenced, yet uncharacterized genomes. Thus, our study is a significant advance in identification of GIs and horizontally transferred genes. SSG-LUGIA is available as an open source software at https://nibtehaz.github.io/SSG-LUGIA/.

Experimental approaches to tracking mobile genetic elements in microbial communities

Horizontal gene transfer is an important mechanism of microbial evolution and is often driven by the movement of mobile genetic elements between cells. Due to the fact that microbes live within communities, various mechanisms of horizontal gene transfer and types of mobile elements can co-occur. However, the ways in which horizontal gene transfer impacts and is impacted by communities containing diverse mobile elements has been challenging to address. Thus, the field would benefit from incorporating community-level information and novel approaches alongside existing methods. Emerging technologies for tracking mobile elements and assigning them to host organisms provide promise for understanding the web of potential DNA transfers in diverse microbial communities more comprehensively. Compared to existing experimental approaches, chromosome conformation capture and methylome analyses have the potential to simultaneously study various types of mobile elements and their associated hosts. We also briefly discuss how fermented food microbiomes, given their experimental tractability and moderate species complexity, make ideal models to which to apply the techniques discussed herein and how they can be used to address outstanding questions in the field of horizontal gene transfer in microbial communities.

Microbial genomic island discovery, visualization and analysis

Horizontal gene transfer (also called lateral gene transfer) is a major mechanism for microbial genome evolution, enabling rapid adaptation and survival in specific niches. Genomic islands (GIs), commonly defined as clusters of bacterial or archaeal genes of probable horizontal origin, are of particular medical, environmental and/or industrial interest, as they disproportionately encode virulence factors and some antimicrobial resistance genes and may harbor entire metabolic pathways that confer a specific adaptation (solvent resistance, symbiosis properties, etc). As large-scale analyses of microbial genomes increases, such as for genomic epidemiology investigations of infectious disease outbreaks in public health, there is increased appreciation of the need to accurately predict and track GIs. Over the past decade, numerous computational tools have been developed to tackle the challenges inherent in accurate GI prediction. We review here the main types of GI prediction methods and discuss their advantages and limitations for a routine analysis of microbial genomes in this era of rapid whole-genome sequencing. An assessment is provided of 20 GI prediction software methods that use sequence-composition bias to identify the GIs, using a reference GI data set from 104 genomes obtained using an independent comparative genomics approach. Finally, we present guidelines to assist researchers in effectively identifying these key genomic regions.

New candidates for regulated gene integrity revealed through precise mapping of integrative genetic elements

Integrative genetic elements (IGEs) are mobile multigene DNA units that integrate into and excise from host bacterial genomes. Each IGE usually targets a specific site within a conserved host gene, integrating in a manner that preserves target gene function. However, a small number of bacterial genes are known to be inactivated upon IGE integration and reactivated upon excision, regulating phenotypes of virulence, mutation rate, and terminal differentiation in multicellular bacteria. The list of regulated gene integrity (RGI) cases has been slow-growing because IGEs have been challenging to precisely and comprehensively locate in genomes. We present software (TIGER) that maps IGEs with unprecedented precision and without attB site bias. TIGER uses a comparative genomic, ping-pong BLAST approach, based on the principle that the IGE integration module (i.e. its int-attP region) is cohesive. The resultant IGEs from 2168 genomes, along with integrase phylogenetic analysis and gene inactivation tests, revealed 19 new cases of genes whose integrity is regulated by IGEs (including dut, eccCa1, gntT, hrpB, merA, ompN, prkA, tqsA, traG, yifB, yfaT and ynfE), as well as recovering previously known cases (in sigK, spsM, comK, mlrA and hlb genes). It also recovered known clades of site-promiscuous integrases and identified possible new ones.

VRprofile: gene-cluster-detection-based profiling of virulence and antibiotic resistance traits encoded within genome sequences of pathogenic bacteria

VRprofile is a Web server that facilitates rapid investigation of virulence and antibiotic resistance genes, as well as extends these trait transfer-related genetic contexts, in newly sequenced pathogenic bacterial genomes. The used backend database MobilomeDB was firstly built on sets of known gene cluster loci of bacterial type III/IV/VI/VII secretion systems and mobile genetic elements, including integrative and conjugative elements, prophages, class I integrons, IS elements and pathogenicity/antibiotic resistance islands. VRprofile is thus able to co-localize the homologs of these conserved gene clusters using HMMer or BLASTp searches. With the integration of the homologous gene cluster search module with a sequence composition module, VRprofile has exhibited better performance for island-like region predictions than the other widely used methods. In addition, VRprofile also provides an integrated Web interface for aligning and visualizing identified gene clusters with MobilomeDB-archived gene clusters, or a variety set of bacterial genomes. VRprofile might contribute to meet the increasing demands of re-annotations of bacterial variable regions, and aid in the real-time definitions of disease-relevant gene clusters in pathogenic bacteria of interest. VRprofile is freely available at http://bioinfo-mml.sjtu.edu.cn/VRprofile.

Comparative Analysis of Genomic Island Prediction Tools

Tools for genomic island prediction use strategies for genomic comparison analysis and sequence composition analysis. The goal of comparative analysis is to identify unique regions in the genomes of related organisms, whereas sequence composition analysis evaluates and relates the composition of specific regions with other regions in the genome. The goal of this study was to qualitatively and quantitatively evaluate extant genomic island predictors. We chose tools reported to produce significant results using sequence composition prediction, comparative genomics, and hybrid genomics methods. To maintain diversity, the tools were applied to eight complete genomes of organisms with distinct characteristics and belonging to different families. Escherichia coli CFT073 was used as a control and considered as the gold standard because its islands were previously curated in vitro. The results of predictions with the gold standard were manually curated, and the content and characteristics of each predicted island were analyzed. For other organisms, we created GenBank (GBK) files using Artemis software for each predicted island. We copied only the amino acid sequences from the coding sequence and constructed a multi-FASTA file for each predictor. We used BLASTp to compare all results and generate hits to evaluate similarities and differences among the predictions. Comparison of the results with the gold standard revealed that GIPSy produced the best results, covering ~91% of the composition and regions of the islands, followed by Alien Hunter (81%), IslandViewer (47.8%), Predict Bias (31%), GI Hunter (17%), and Zisland Explorer (16%). The tools with the best results in the analyzes of the set of organisms were the same ones that presented better performance in the tests with the gold standard.

Evolution of the U.S. Biological Select Agent Rathayibacter toxicus

Rathayibacter toxicus is a toxin-producing species found in Australia and is often fatal to grazing animals. The threat of introduction of the species into the United States led to its inclusion in the Federal Select Agent Program, which makes R. toxicus a highly regulated species. This work provides novel insights into the evolution of R. toxicus. R. toxicus is the only species in the genus to have acquired a CRISPR adaptive immune system to protect against bacteriophages. Results suggest that coexistence with the bacteriophage NCPPB3778 led to the massive shrinkage of the R. toxicus genome, species divergence, and the maintenance of low genetic diversity in extant bacterial groups. This work contributes to an understanding of the evolution and ecology of an agriculturally important species of bacteria.

Rathayibacter toxicus is a species of Gram-positive, corynetoxin-producing bacteria that causes annual ryegrass toxicity, a disease often fatal to grazing animals. A phylogenomic approach was employed to model the evolution of R. toxicus to explain the low genetic diversity observed among isolates collected during a 30-year period of sampling in three regions of Australia, gain insight into the taxonomy of Rathayibacter, and provide a framework for studying these bacteria. Analyses of a data set of more than 100 sequenced Rathayibacter genomes indicated that Rathayibacter forms nine species-level groups. R. toxicus is the most genetically distant, and evidence suggested that this species experienced a dramatic event in its evolution. Its genome is significantly reduced in size but is colinear to those of sister species. Moreover, R. toxicus has low intergroup genomic diversity and almost no intragroup genomic diversity between ecologically separated isolates. R. toxicus is the only species of the genus that encodes a clustered regularly interspaced short palindromic repeat (CRISPR) locus and that is known to host a bacteriophage parasite. The spacers, which represent a chronological history of infections, were characterized for information on past events. We propose a three-stage process that emphasizes the importance of the bacteriophage and CRISPR in the genome reduction and low genetic diversity of the R. toxicus species.

GI-Cluster: Detecting genomic islands via consensus clustering on multiple features

The accurate detection of genomic islands (GIs) in microbial genomes is important for both evolutionary study and medical research, because GIs may promote genome evolution and contain genes involved in pathogenesis. Various computational methods have been developed to predict GIs over the years. However, most of them cannot make full use of GI-associated features to achieve desirable performance. Additionally, many methods cannot be directly applied to newly sequenced genomes. We develop a new method called GI-Cluster, which provides an effective way to integrate multiple GI-related features via consensus clustering. GI-Cluster does not require training datasets or existing genome annotations, but it can still achieve comparable or better performance than supervised learning methods in comprehensive evaluations. Moreover, GI-Cluster is widely applicable, either to complete and incomplete genomes or to initial GI predictions from other programs. GI-Cluster also provides plots to visualize the distribution of predicted GIs and related features. GI-Cluster is available at https://github.com/icelu/GI_Cluster.

xenoGI: reconstructing the history of genomic island insertions in clades of closely related bacteria

Background

Genomic islands play an important role in microbial genome evolution, providing a mechanism for strains to adapt to new ecological conditions. A variety of computational methods, both genome-composition based and comparative, have been developed to identify them. Some of these methods are explicitly designed to work in single strains, while others make use of multiple strains. In general, existing methods do not identify islands in the context of the phylogeny in which they evolved. Even multiple strain approaches are best suited to identifying genomic islands that are present in one strain but absent in others. They do not automatically recognize islands which are shared between some strains in the clade or determine the branch on which these islands inserted within the phylogenetic tree.

Results

We have developed a software package, xenoGI, that identifies genomic islands and maps their origin within a clade of closely related bacteria, determining which branch they inserted on. It takes as input a set of sequenced genomes and a tree specifying their phylogenetic relationships. Making heavy use of synteny information, the package builds gene families in a species-tree-aware way, and then attempts to combine into islands those families whose members are adjacent and whose most recent common ancestor is shared. The package provides a variety of text-based analysis functions, as well as the ability to export genomic islands into formats suitable for viewing in a genome browser. We demonstrate the capabilities of the package with several examples from enteric bacteria, including an examination of the evolution of the acid fitness island in the genus Escherichia. In addition we use output from simulations and a set of known genomic islands from the literature to show that xenoGI can accurately identify genomic islands and place them on a phylogenetic tree.

Conclusions

xenoGI is an effective tool for studying the history of genomic island insertions in a clade of microbes. It identifies genomic islands, and determines which branch they inserted on within the phylogenetic tree for the clade. Such information is valuable because it helps us understand the adaptive path that has produced living species.

Electronic supplementary material

The online version of this article (10.1186/s12859-018-2038-0) contains supplementary material, which is available to authorized users.

Computational methods for predicting genomic islands in microbial genomes

Clusters of genes acquired by lateral gene transfer in microbial genomes, are broadly referred to as genomic islands (GIs). GIs often carry genes important for genome evolution and adaptation to niches, such as genes involved in pathogenesis and antibiotic resistance. Therefore, GI prediction has gradually become an important part of microbial genome analysis. Despite inherent difficulties in identifying GIs, many computational methods have been developed and show good performance. In this mini-review, we first summarize the general challenges in predicting GIs. Then we group existing GI detection methods by their input, briefly describe representative methods in each group, and discuss their advantages as well as limitations. Finally, we look into the potential improvements for better GI prediction.