Robust high throughput prokaryote de novo assembly and improvement pipeline for Illumina data

Abstract

The rapidly reducing cost of bacterial genome sequencing has lead to its routine use in large scale microbial analysis. Though mapping approaches can be used to find differences relative to the reference, many bacteria are subject to constant evolutionary pressures resulting in events such as the loss and gain of mobile genetic elements, horizontal gene transfer through recombination and genomic rearrangements. De novo assembly is the reconstruction of the underlying genome sequence, an essential step to understanding bacterial genome diversity. Here we present a high throughput bacterial assembly and improvement pipeline that has been used to generate nearly 20,000 draft genome assemblies in public databases. We demonstrate its performance on a public data set of 9,404 genomes. We find all the genes used in MLST schema present in 99.6% of assembled genomes. When tested on low, neutral and high GC organisms, more than 94% of genes were present and completely intact. The pipeline has proven to be scalable and robust with a wide variety of datasets without requiring human intervention. All of the software is available on GitHub under the GNU GPL open source license.\n\nDATA SUMMARYO_LIThe assembly pipeline software is available from Github under the GNU GPL open source license; (url - https://github.com/sanger-pathogens/vr-codebase)\nC_LIO_LIThe assembly improvement software is available from Github under the GNU GPL open source license; (url - https://github.com/sanger-pathogens/assembly_improvement)\nC_LIO_LIAccession numbers for 9,404 assemblies are provided in the supplementary material.\nC_LIO_LIThe Bordetella pertussis sample has sample accession ERS1058649, sequencing reads accession ERR1274624 and assembly accessions FJMX01000001-FJMX01000249.\nC_LIO_LIThe Salmonella enterica subsp. enterica serovar Pullorum sample has sample accession ERS1058652, sequencing reads accession ERR1274625 and assembly accession FJMV01000001-FJMV01000026.\nC_LIO_LIThe Staphylococcus aureus sample has sample accession ERS1058648, sequencing reads accession ERR1274626 and assembly accessions FJMW01000001-FJMW01000040.\nC_LI\n\nI/We confirm all supporting data, code and protocols have been provided within the article or through supplementary data files.{ballotcheck}\n\nIMPACT STATEMENTThe pipeline described in this paper has been used to assemble and annotate 30% of all bacterial genome assemblies in GenBank (18,080 out of 59,536, accessed 16/2/16). The automated generation of de novo assemblies is a critical step to explore bacterial genome diversity. MLST genes are found in 99.6% of cases, making it at least as good as existing typing methods. In the test genomes we present, more than 94% of genes are correctly assembled into intact reading frames.