Accurate and fast graph-based pangenome annotation and clustering with ggCaller

Abstract

Bacterial genomes differ in both gene content and sequence mutations, which can cause important clinical phenotypic differences such as vaccine escape or antimicrobial resistance. To identify and quantify important variants, all genes within a population must be predicted, functionally annotated and clustered, representing the 'pangenome'. Despite the volume of genome data available, gene prediction and annotation are currently conducted in isolation on individual genomes, which is computationally inefficient and frequently inconsistent across genomes. Here, we introduce the open-source software graph-gene-caller (ggCaller; https://github.com/samhorsfield96/ggCaller). ggCaller combines gene prediction, functional annotation and clustering into a single step using population-wide de Bruijn Graphs, removing redundancy in gene annotation, and resulting in more accurate gene predictions and orthologue clustering. We applied ggCaller to simulated and real-world bacterial genome datasets, comparing it to current state-of-the-art tools. ggCaller is ~50x faster with equivalent or greater accuracy, particularly in datasets with complex sources of error, such as assembly contamination or fragmentation. ggCaller is also an important extension to bacterial genome-wide association studies, enabling querying of annotated graphs for functional analyses. We highlight this application by functionally annotating DNA sequences with significant associations to tetracycline and macrolide resistance in Streptococcus pneumoniae, identifying key resistance determinants that were missed when using only a single reference genome. ggCaller is a novel bacterial genome analysis tool with applications in bacterial epidemiology and evolutionary study.

Improved interpretability of bacterial genome-wide associations using gene cluster centric k-mers

“…An ideal dataset would then have harmonized gene annotations across isolates. Recent advances in pangenome-aware gene calling and pangenome estimation might then be more appropriate; the recently described ggCaller 15 algorithm's output is in fact compatible with panfeed.…”

“…panfeed iterates over the gene presence/absence matrix produced by software such as Roary 20 , panaroo 21 or ggCaller 15 , specifically the "gene_presence_absence.csv" file. For each gene cluster, the full nucleotide sequence of the gene is retrieved from the input GFF3 file of each of the input samples, optionally including sequences upstream and downstream of the start and stop codon, respectively.…”

“…This makes the use of reference genomes, as is typically done in human-based studies, less appealing, since the large genomic heterogeneity cannot be captured by using a single reference genome; this is especially true for the extreme case of accessory genes not encoded in the chosen reference 12 . An approach that mitigates this reference bias is the use of pangenome graphs, which are often constructed using de Bruijn graphs (DBG) via k-mers 13 , unitigs 14 or gene clusters 15 as the smallest construction units.…”

“…Encoding genetic variants using k-mers in the context of each gene cluster of the pangenome would therefore allow to account for both reference bias, since encoding variants using k-mers does not require the use of a reference, and for improved interpretability, since panfeed can retain the absolute and relative position of each k-mer in each input genome, thus allowing a fine grained mapping of associated variants. panfeed uses as input a genes' presence/absence matrix (as the one given by Roary 20 , panaroo 21 or ggCaller 15 ) to create cluster specific k-mer sets and k-mer presence/absence patterns, which can be fed directly into bacterial GWAS software such as pyseer 22 . The resulting association table can easily be correlated to the gene clusters, which enables fast interpretation and visualization of variants occurring in both core and accessory genomes.…”

PanGraph: scalable bacterial pan-genome graph construction

“…Coloured and compacted generalizations of the de Bruijn graph-based assemblers have been successively used to build graphs from large sequence sets [11,12], with tools existing to build graphs containing thousands of isolates [13]. These graphs have been shown to be useful in problems such as the detection of the core genome [14] or the improvement of gene prediction, annotation and clustering [15]. However, in these methods the graph structure encodes, at the same time, nucleotide-level variation and largescale structural rearrangements.…”