GFFview: A Web Server for Parsing and Visualizing Annotation Information of Eukaryotic Genome

1. Introduction

Owing to rapid advance of high-throughput sequencing technology, RNA sequencing (RNA-seq) has significantly revolutionized the transcriptome analysis in eukaryotes (Wang et al., 2009). RNA-seq could be efficiently used for quantifying gene expression, identifying post-transcriptional processing, and annotating gene structure (Mortazavi et al., 2008; Wang et al., 2010; Roberts et al., 2011). Along with the increasing understanding of gene transcriptions, more and more eukaryotic genomes have been extensively annotated (Besemer et al., 2001; Roberts et al., 2011). Purposes for annotating genome mainly include description of the exon–intron boundaries, coding and noncoding regions, alternative splicing isoforms, alternative polyadenylation, transcription orientation, and biotypes of gene products.

For a representative eukaryote species, genome sequence would be more than 1500 Mb in size and is associated with a large amount of annotation information (Huang et al., 2016). To efficiently store and share such enormous information, therefore, a standard General Feature Format (GFF) was proposed and designed as tab-delimited text file (http://gmod.org/wiki/GFF3), which contains both mandatory and optional fields. The GFF has been widely used by not only the popular public databases (Derrien et al., 2012; Yates et al., 2016) but also a large number of bioinformatic tools (Li and Dewey, 2011). Despite this, GFF file would be very large in size, and for which it is difficult to manipulate by biologists who have no bioinformatic skill. For example, the plain text file in GFF for recording annotation information of human genome is 1343 Mbp in size and has 2,575,499 line records in total for current release of Ensembl (Yates et al., 2016). In addition to GFF, the Gene Transfer Format (GTF) is also a similar but more specific format for recording information in relation to gene structure. However, there is no essential difference between GFF and GTF, both of which could be transformed from each other. In this study, we develop a web server for parsing and visualizing annotation information of eukaryotic genome as being stored in GFF or GTF, which could be easily used for schematically generating whole-transcriptome landscape.

2. Design and Implementation of GFFview

2.1. Parsing of GFF or GTF file

The initial step of GFFview is to parse annotation information of genome in GFF or GTF (Fig. 1A). First, we extract all features with biological types as being labeled as gene, transcript/messenger RNA, exon, or UTR (Untranslated Regions). Second, chromosomal locations for each feature are individually determined. Finally, different annotation information, such as the gene structure and isoforms, could be fully constructed on the basis of these recorded locations. Because we ignore other features of genome annotation such as CDS (Coding Sequence), start code, and end code, the computational memory usage would be largely reduced.

OPEN IN VIEWER

FIG. 1.

Schematic view of GFFview workflow. (A) Screenshot of the GFFview. (B) Operation procedures to use GFFview are shown. GFF, General Feature Format; GTF, Gene Transfer Format; UTR, Untranslated Regions.

3. Result and Discusion

We prepared a demo data containing complete annotation information of human chromosomes in GTF, which totally has 1,713,977 line records (Yates et al., 2016). Six indices were generated with GFFview and individually shown in Figure 2. Based on these results, we can easily obtain overall landscape at the whole-transcriptome level, which would be very useful when checking quality and difference of the de novo assembled transcriptome in RNA-seq studies.

OPEN IN VIEWER

FIG. 2.

GFFview images showing profile of human genome annotation in GTF file. We demonstrated the gene density within each chromosome (A), gene length distribution (B), and number of transcripts per gene (C). In addition, the information of length distribution, exon number, and annotated UTR(s) of all transcripts is shown on density curve (D), bar graphic (E), and bin chart (F), respectively.

In contrast to the existing softwares such as Integrative Genomics Viewer (Thorvaldsdóttir et al., 2013) and gff2ps (Abril and Guigó, 2000), which tend to focus on structural details of specific gene, GFFview was developed to provide an overall view for describing annotation information of genome. Of course, functions provided by GFFview could be implemented with R package and Python library. However, the web-based application of GFFview would largely facilitate use by biologists. One potential issue for using GFFview is that the data need to be uploaded, which would be a time-consuming and expensive task when the file is considerable in size. Furthermore, a user can delete the uploaded file thoroughly or completely, the web server will not store any data in private.