Prokaryotic Contig Annotation Pipeline Server: Web Application for a Prokaryotic Genome Annotation Pipeline Based on the Shiny App Package

1. Introduction

As prokaryotic genome sequencing becomes increasingly cheaper and commonplace, there is a greater demand for a more easy-to-use genome analysis tool for individual laboratories. There are generally two types of prokaryotic genome annotation pipelines available for users: (1) a remote web server-based service or (2) a locally installable stand-alone package. The web server-based pipeline, represented by RAST (Rapid Annotations using Subsystems Technology) (Aziz et al., 2008), provides the genome annotation on a remote web server through sequence submission and alerts users by e-mail. The web server-based pipelines use predetermined processes that are initiated by data submission, which can raise concerns about data privacy. In contrast, use of a locally installed stand-alone pipeline, represented by Prokka (Seemann, 2014), avoids issues of data transfer and security and can be adjusted by the users. However, the installation requires relevant computing environments.

To complement these issues and provide a more generally accessible genome annotation pipeline, we developed a web-based application (webApp) termed P-CAPS (Prokaryotic Contig Annotation Pipeline Server). The pipeline is managed under the ShinyR package, a webApp framework for R. P-CAPS can be easily deployed on the local Linux system equipped with various open-source bioinformatics tools for genome annotation. Users can also launch P-CAPS as their own webApp service in a Shiny server. The pipeline is maintained with an R script language without complete knowledge about HTML or JavaScript. P-CAPS lets users embed additional R scripts for their own analyses. Because the pipeline is handled directly in the web browser, users can plug external programs such as JBrowse (Skinner et al., 2009) into P-CAPS. A graphical overview written in R markdown is provided as a summary report.

2. Description

2.3. Output

P-CAPS provides the annotation result in various standard file formats such as FASTA, GFF3, and GBFF (GBK), along with a graphical summary report written in R markdown. The summary report presents basic statistics of genome annotation such as sequence length, GC ratio, and annotated feature counts. The annotation completeness by Benchmarking Universal Single-Copy Ortholog (BUSCO) (Simão et al., 2015) and the functional categorization by KO are included in the summary report (Fig. 1).

OPEN IN VIEWER

FIG. 1.

Example of a graphical summary report from P-CAPS. The summary report is created with R markdown and is downloadable in DOCX, PDF, and HTML formats. P-CAPS, Prokaryotic Contig Annotation Pipeline Server; rRNA, ribosomal RNA; tRNA, transfer RNA.

3. Results

To evaluate the performance of P-CAPS, we benchmarked the package against publicly available annotation programs (RAST and Prokka) using NCBI (National Center for Biotechnology Information) reference genomes from various taxa: Escherichia coli str. K-12 substr. MG1655 (GCA_000005845.2), Lactobacillus plantarum WCFS1 (GCA_000203855.3), Bacillus subtilis subsp. subtilis str. 168 (GCA_000009045.1), Eubacterium limosum KIST612 (GCA_000152245.2), and Vibrio sp. EJY3 (GCA_000241385.1). P-CAPS showed comparable performance to the other programs for predicting major genomic features such as the CDSs, tRNAs, and rRNAs (Table 1). We checked the annotation quality according to the completeness of genome annotation using BUSCO (Simão et al., 2015). The annotations from P-CAPS and Prokka showed near completeness of the BUSCO checkup over the tested NCBI RefSeq genomes (Table 2). RAST missed a few CDSs from incongruent start sites of the RefSeq, which was due to short and fragmented protein sequence prediction (Fig. 2). The functionality of P-CAPS as a web application with a Shiny server is tested in the server of Korea University (http://panflam.korea.ac.kr/pcaps/mainpage).

OPEN IN VIEWER

FIG. 2.

Example of the missed and fragmented BUSCOs annotated by RAST. The results were not in accordance with the annotation from P-CAPS and other pipelines. The RAST annotation showed (a) one missed BUSCO and (b) five fragmented BUSCOs on the Escherichia coli genome. BUSCOs, Benchmarking Universal Single-Copy Orthologs; CDS, coding sequence; NCBI, National Center for Biotechnology Information; RAST, Rapid Annotations using Subsystems Technology.