Reticulate Representation of Evolutionary and Functional Relationships between Phage Genomes

Abstract

Bacteriophage genomes show pervasive mosaicism, indicating the importance of horizontal gene exchange in their evolution. Phage genomes represent unique combinations of modules, each of them with a different phylogenetic history. The traditional classification, based on a variety of criteria such as nucleic acid type (single/double-stranded DNA/RNA), morphology, and host range, appeared inconsistent with sequence analyses. With the genomic era, an ever increasing number of sequenced phages cannot be classified, in part due to a lack of morphological information and in part to the intrinsic incapability of tree-based methods to efficiently deal with mosaicism. This problem led some virologists to call for a moratorium on the creation of additional taxa in the order Caudovirales, in order to let virologists discuss classification schemes that might better suit phage evolution. In this context, we propose a framework for a reticulate classification of phages based on gene content. Starting from gene families, we built a weighted graph, where nodes represent phages and edges represent phage-phage similarities in terms of shared genes. We then apply various measures of graph topology to analyze the resulting graph. Most double-stranded DNA phages are found in a single component. The values of the clustering coefficient and closeness distinguish temperate from virulent phages, whereas chimeric phages are characterized by a high betweenness coefficient. We apply a 2-step clustering method to this graph to generate a reticulate classification of phages: Each phage is associated with a membership vector, which quantitatively characterizes its membership to the set of clusters. Furthermore, we cluster genes based on their "phylogenetic profiles" to define "evolutionary cohesive modules." In virulent phages, evolutionary modules span several functional categories, whereas in temperate phages they correspond better to functional modules. Moreover, despite the fact that modules only cover a fraction of all phage genes, phage groups can be distinguished by their different combination of modules, serving the bases for a higher level reticulate classification. These 2 classification schemes provide an automatic and dynamic way of representing the relationships within the phage population and can be extended to include newly sequenced phage genomes, as well as other types of genetic elements.

Network analyses structure genetic diversity in independent genetic worlds

“…S1) when we investigated the network reconstructed with sequences presenting at least 40% of sequence identity. This is either because of the methodological limit of the method that cannot account for edge weights, or, as it is in agreement with the literature (15)(16)(17)(18), it might be explained if, within the limits of one type of vehicle, DNA molecules obey complex rules of transfers and losses. Yet, the lack of a strong underlying hierarchical structure does not mean that understanding the processes that led to its complexity is out of reach.…”

“…To obtain a robust estimate of these centralities, they were computed for the members of the GCC only. As different vehicles do not harbor the same number of DNA families, we corrected for the bias in genome size and kept only edges corresponding to a significant number of shared DNA families (18).…”

Ecogenomics and potential biogeochemical impacts of globally abundant ocean viruses

“…vContact (https://bitbucket.org/MAVERICLab/ vcontact) was then used to calculate a similarity score between every pair of genome and/or contigs based on the number shared of PCs between the two sequences (as in 8,9 ), and then compute a MCL clustering of the genomes/contigs based on these similarity scores (thresholds of 1 on similarity score, MCL inflation of 2). The resulting viral clusters (or VCs, clusters including ≥2 contigs and/or genomes), consistent with a clustering based on whole-genome BLAST comparison, corresponded to approximately genus-level taxonomy, with rare cases closer to subfamily-level taxonomy (Extended Data Fig.…”

“…1) using shared gene content information and network analytics 8 . This method starts from genome fragments (≥10kb) and results in VCs approximately equivalent to known bacterial and archaeal virus genera 8,9 (see also Supplementary Text, Extended Data Table 3 for comparison with alternative classification methods). Combining the 15,222 viral populations identified here with the genomes and genome fragments of another 15,929 publicly available bacterial and archaeal viruses generated a total of 1,259 VCs.…”

iVirus: facilitating new insights in viral ecology with software and community datasets imbedded in a cyberinfrastructure

“…This is critical as viruses lack a universal gene marker (Edwards and Rohwer, 2005) and o0.1% of viruses in natural environments are represented in public databases (Brum et al, 2015), which necessitates new approaches to taxonomically classify surveyed viral genomes. Inspired by algorithms to detect prophage in microbial genomes (Lima-Mendez et al, 2008), vContact clusters contigs by their PC profiles (note: see preferred method for PC generation as vContact-PCs, but the user could generate PCs however they prefer). Reference sequences and their taxonomic lineages can be seeded within the analysis to improve clustering and taxonomic predictions.…”