Abstract
BackgroundBacterial gene loss and acquisition is a well-known phenomenon which contributes to bacterial adaptation through changes in important phenotypes such as virulence, antibiotic resistance and metabolic capability. While advances in DNA sequencing have accelerated our ability to generate short genome sequence reads to disentangle phenotypic changes caused by gene loss and acquisition, the short-read genome sequencing often results in fragmented genome assemblies as a basis for identification of gene loss and acquisition events. However, sensitive and precise determination of gene content change for fragmented genome assemblies remains challenging as analysis needs to account for cases when only a fragment of the gene is assembled or when the gene assembly is split in more than one contig.ResultsWe developed GenAPI, a command-line tool that is designed to compare the gene content of bacterial genomes for which only fragmented genome assemblies are available. GenAPI, unlike other available tools of similar purpose, accounts for imperfections in sequencing and assembly, and aims to compensate for them. We tested the performance of GenAPI on three different datasets to show that GenAPI has a high sensitivity while it maintains precision when dealing with partly assembled genes in both simulated and real datasets. Furthermore, we benchmarked the performance of GenAPI with six popular tools for gene presence-absence identification.ConclusionsOur developed bioinformatics tool, called GenAPI, has the same precision and recall rates when analyzing complete genome sequences as the other tools of the same purpose; however, GenAPI’s performance is markedly better on fragmented genome assemblies.
Highlights
Bacterial gene loss and acquisition is a well-known phenomenon which contributes to bacterial adaptation through changes in important phenotypes such as virulence, antibiotic resistance and metabolic capability
We evaluated the performance of GenAPI on three test datasets with genome assemblies; two simulated datasets with in silico introduced variation in gene content (P. aeruginosa and Salmonella typhi [4] datasets, respectively), and one real dataset with known gene deletions (Escherichia coli experiment [19])
Genes which were correctly predicted to be absent were defined as true positive (TP) gene absence; genes incorrectly predicted to be present were defined as false negative (FN) gene absence; and genes incorrectly predicted to be absent were defined as false positive (FP) gene absence
Summary
Bacterial gene loss and acquisition is a well-known phenomenon which contributes to bacterial adaptation through changes in important phenotypes such as virulence, antibiotic resistance and metabolic capability. E.g. PanSeq [2], are based on alignment of the query genome sequence to a reference genome to test for the presence-absence of genes that are present in the given reference This approach is limited knowing that prophage and plasmid genes constitute a major part of the variable bacterial genetic content [3]. Other tools, such as Roary [4], SaturnV [5], PanDelos [6], panX [7], BPGA [8] and EDGAR [9] construct a pan-genome from inputted genome assemblies and determine the gene set that is present in each of the genome assemblies. There is a need for analytical tools that are designed to account for highly fragmented assemblies (often the product of de novo genome assemblies based on short-read sequence data where genome assemblies have tens to hundreds of contigs) that else would result in a large number of false calls for gene being absent in the assembly
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
