GrapeTree: visualization of core genomic relationships among 100,000 bacterial pathogens.
Zhou Z, Alikhan NF, Sergeant MJ, Luhmann N, Vaz C, Francisco AP, Carriço JA, Achtman M. · 2018
Read GrapeTree: visualization of core genomic relationships among 100,000 bacterial pathogens. by Zhou Z, Alikhan NF, Sergeant MJ, Luhmann N, Vaz C, Francisco AP, Carriço JA, Achtman M. (2018) 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-pmc6120633.txt.
Introduction
Zhou and co-authors present a software article in Genome Research on GrapeTree, a tool for viewing links among bacterial genomes. Health teams track germs by comparing gene profiles. Databases now hold tens of thousands of strains. Old methods strain to draw trees that big.
GrapeTree draws family trees for up to 100,000 bacterial pathogens. A layout method keeps large trees easy to explore in a web browser. A new tree-building method handles data with many gaps. The tool works alone or inside EnteroBase, a pathogen database. Readers get a picture of how outbreak tracking scales to the genome era.
Catalog record
| CREATOR | Zhou Z, Alikhan NF, Sergeant MJ, Luhmann N, Vaz C, Francisco AP, Carriço JA, Achtman M. |
|---|---|
| YEAR | 2018 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 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:
GrapeTree: visualization of core genomic relationships among 100,000 bacterial pathogens By Zhemin Zhou, Nabil-Fareed Alikhan, Martin J Sergeant, Nina Luhmann, Cátia Vaz, Alexandre P Francisco, João André Carriço, Mark Achtman Published in: Genome Research DOI: 10.1101/gr.232397.117 ABSTRACT Abstract Current methods struggle to reconstruct and visualize the genomic relationships of large numbers of bacterial genomes. GrapeTree facilitates the analyses of large numbers of allelic profiles by a static “GrapeTree Layout” algorithm that supports interactive visualizations of large trees within a web browser window. GrapeTree also implements a novel minimum spanning tree algorithm (MSTree V2) to reconstruct genetic relationships despite high levels of missing data. GrapeTree is a stand-alone package for investigating phylogenetic trees plus associated metadata and is also integrated into EnteroBase to facilitate cutting edge navigation of genomic relationships among bacterial pathogens. Legacy MLST (multilocus sequence typing) based on seven housekeeping genes was introduced 20 years ago (Maiden et al. 1998) and is now routinely used for the characterization of numerous bacterial pathogens (Jolley and Maiden 2014). MLST assigns distinct integer numbers to each unique sequence (allele) and a distinct integer number, the sequence type (ST), to each unique combination of allelic integers. Unrelated STs share few alleles or none at all. In contrast, STs that share all but one or two alleles are considered to be strongly related even if the differing alleles contain multiple SNPs due to recombination. The largest legacy MLST databases contain data on ≥60,000 bacterial strains (https://pubmlst.org/databases.shtml). In order to support epidemiological tracking of transmission networks and disease control, the resolution achieved by MLST was recently expanded to encompass more than seven gene fragments. Expanded MLST schemes can include all 53 genes encoding ribosomal proteins (rMLST) (Jolley et al. 2012), thousands of core genes that are present in most isolates of a species or genus (core genome MLST, cgMLST) (Mellmann et al. 2011; Maiden et al. 2013; Moura et al. 2016), or even all the genes in the entire genome (whole genome MLST, wgMLST) (Nadon et al. 2017). We have recently developed EnteroBase (https://enterobase.warwick.ac.uk), a genotyping website for selected enteric pathogens (Alikhan et al. 2018). EnteroBase automatically assembles Illumina short reads into contigs and assigns the assembled sequences to MLST alleles and STs at all levels of resolution from legacy MLST through to wgMLST. EnteroBase performs these operations for short reads that are in the public domain or uploaded by users. In April 2018, EnteroBase contained ∼130,000 Salmonella genomes and >65,000 Escherichia genomes, and the numbers of sets of Illumina short reads in the public domain continues to grow rapidly (Alikhan et al. 2018). A driving force behind developing such large databases is to facilitate our understanding of epidemiological and population genetic phenomena among isolates from distinct geographical sources and over extended time scales. Initially, the genetic relationships of legacy STs were represented by phylograms based on hierarchical clustering methods, an approach which can be very useful for visualizing deeper branching structures. Phylograms may, however, be problematic for the presentation of large numbers of genotypes because each genotype is represented by a unique branch, even when multiple genotypes are identical. An example of this problem arises when visualizing the allelic distances between 99,722 Salmonella spp. strains from 3902 legacy MLST STs. The associations between serovars and genetic clades are somewhat difficult to interpret within the default presentation of this phylogram by iTOL (Fig. 1A; Letunic and Bork 2016). Dendrograms generated by other programs (FigTree [v.1.4.3, http://tree.bio.ed.ac.uk/software/figtree/]; Dendroscope [Huson and Scornavacca 2012]) from large data sets were also difficult to interpret. Still other graphical user interfaces were unable to even depict this large number of items, including PHYLOViZ 2.0 (Nascimento et al. 2017), SplitsTree4 (Huson and Bryant 2006), EvolView (He et al. 2016), Microreact (Argimon et al. 2016), TreeDyn (Chevenet et al. 2006), TreeView (Page 1996), and Phandango (Hadfield et al. 2018). Similarly, handling more than 5000 genomes presents problems for de novo sequence-based SNP comparisons (Mazariegos-Canellas et al. 2017), and trees based on phylogenetic algorithms are difficult to comprehend when they contain large numbers of nodes (Fig. 1B,D). [Figure: Visualization of 3902 legacy Salmonella MLST STs from 99,722 genomic assemblies in EnteroBase (Alikhan et al. 2018) by a phylogram versus a minimum spanning tree. (A,B) iTol (Letunic and Bork 2016) visualization of genetic relationships. Nodes at the ends of the terminal edges represent each of the 99,722 genomic assemblies. (C,D) GrapeTree visualization of genetic clusters. Nodes represent each of the 3902 STs, with diameters scaled to the number of assemblies. In C, edges between nodes mark allelic distances of 1–2 of the seven loci. (A,C) Representation of a minimum spanning tree generated by MSTree V2 in Newick format. (B,D) Representation of a neighbor-joining tree generated by RapidNJ (Simonsen et al. 2011) in Newick format. Color codes for the 60 most common serovars are indicated in the central key legend and used to color branches plus wedges in an external circle (A,B) or individual nodes (C,D). Interactive versions of the trees can be found at (A) http://bit.ly/2qH06jp, (B) https://bit.ly/2mDOpbS, (C) http://bit.ly/2H69dkG, and (D) https://bit.ly/2LG62Tl.] An alternative to phylograms is minimum spanning trees, which have less demanding graphical requirements because they map clusters of related nodes in 2D space (Francisco et al. 2012; Nascimento et al. 2017). A commercial software program (BioNumerics, Applied Maths) introduced an improved visualization of a minimum spanning tree to microbiologists in the early 1990s, which reduced complexity by grouping isolates with identical STs within single nodes whose diameter reflected the numbers of isolates. A similar visualization was subsequently offered by the noncommercial PHYLOViZ software (Francisco et al. 2009). We have now extended these approaches with GrapeTree, a software package that supports the efficient visualization of minimum spanning trees and phylograms from character data. An initial indication of its capabilities can be gained by comparing the representations of genetic relationships according to legacy MLST data by iTOL (Fig. 1A,B) and GrapeTree (Fig. 1C,D). Calculating minimum spanning trees from legacy MLST is quick and efficient because legacy MLST is based on only seven loci, and allelic calls for each of the seven loci are a prerequisite for calling an ST, i.e., no missing data. As a result, the GrapeTree visualization in Figure 1C took only 1.5 min. However, the cgMLST of Salmonella spans 3002 loci (Alikhan et al. 2018), and STs routinely include low levels of missing data because some cgMLST genes are occasionally deleted or are not identified due to various bioinformatics problems in the assembly of genomes from short reads. As a result, multiple sets of almost identical STs exist in EnteroBase that only differ due to missing data, but each of which is, nevertheless, a unique node in a phylogram because its allelic content differs from those of other STs. As demonstrated below, missing data are also a problem for the classical minimum spanning tree approach (henceforth MSTree) implemented by BioNumerics and goeBURST (Francisco et al. 2009). We have therefore implemented MSTree V2, which is an improved algorithm for generating minimum spanning trees from character sets that contain missing data. Here we present GrapeTree, a web browser application that efficiently reconstructs and visualizes intricate minimum spanning trees together with detailed metadata. RESULTS Overview of GrapeTree features GrapeTree is a fully interactive,