PPanGGOLiN: Depicting microbial diversity via a partitioned pangenome graph

Abstract

The use of comparative genomics for functional, evolutionary, and epidemiological studies requires methods to classify gene families in terms of occurrence in a given species. These methods usually lack multivariate statistical models to infer the partitions and the optimal number of classes and don't account for genome organization. We introduce a graph structure to model pangenomes in which nodes represent gene families and edges represent genomic neighborhood. Our method, named PPanGGOLiN, partitions nodes using an Expectation-Maximization algorithm based on multivariate Bernoulli Mixture Model coupled with a Markov Random Field. This approach takes into account the topology of the graph and the presence/absence of genes in pangenomes to classify gene families into persistent, cloud, and one or several shell partitions. By analyzing the partitioned pangenome graphs of isolate genomes from 439 species and metagenome-assembled genomes from 78 species, we demonstrate that our method is effective in estimating the persistent genome. Interestingly, it shows that the shell genome is a key element to understand genome dynamics, presumably because it reflects how genes present at intermediate frequencies drive adaptation of species, and its proportion in genomes is independent of genome size. The graph-based approach proposed by PPanGGOLiN is useful to depict the overall genomic diversity of thousands of strains in a compact structure and provides an effective basis for very large scale comparative genomics. The software is freely available at https://github.com/labgem/ PPanGGOLiN. PLOS COMPUTATIONAL BIOLOGYPLOS Computational Biology | https://doi.org/10.

Construction of a complete set of Neisseria meningitidis mutants and its use for the phenotypic profiling of this human pathogen

“…Since corresponding proteins are expected to be broadly conserved, we determined the partition of the 391 essential genes between conserved and variable genomes in meningococci. To do this, we first used the recently described PPanGGOLiN software 21 As for similar efforts in other bacteria 10,12,13 , we first selected protein-coding genes to be targeted by systematic mutagenesis, excluding 85 genes (4.1%, highlighted in black in the first pie chart) because they encode transposases of repeated insertion sequences, or correspond to short remnants of truncated genes or cassettes (Supplementary Data 1). We then followed a two-step mutagenesis approach explained in the text and in Supplementary Fig.…”

“…The genome of strain 8013 and a total of 108 complete genomes from the RefSeq database 64 (last accessed June 8th, 2020) (Supplementary Data 5) was used to compute the N. meningitidis pangenome using the PPanGGOLiN software 21 (version 1.1.85). The original annotations of the genomes have been kept in order to compute gene families.…”

“…The original annotations of the genomes have been kept in order to compute gene families. In brief, PPanGGOLiN uses a statistical model to infer the pangenome classes (persistent, shell and cloud) based on both the presence/ absence of gene families and genomic neighbourhood information 21 . The options "-use pseudo" and "-defrag" were used to consider pseudogenes in the gene family computation, and to associate fragmented genes with their original gene family, respectively.…”

Producing polished prokaryotic pangenomes with the Panaroo pipeline

“…Previous approaches which aid in the inference of the pangenome of a collection of bacterial isolates include Roary, OrthoMCL, PanOCT, PIRATE, PanX, PGAP, COGsoft, MultiParanoid, PPanGGoLiN and MetaPGN [4][5][6][7][8][9][10][11][12]. The majority of methods for determining the pangenome tend to make use of one of two similar approaches (see Supplementary Figure 1).…”

“…A final step of some pipelines is to classify the resulting clusters into core and accessory categories based upon their prevalence within the dataset. This is usually done using predefined thresholds; however, more recently model-based extensions to this approach have been suggested [11].…”