Advent of next generation sequencing, together with the organization of several genome projects, made sequencing the genome an affordable task for many organisms. Many gene prediction programs allow the identification of genes within a new genome [1] and the General Feature Format (GFF, proposed by Durbin and Haussler, http://www.sanger.ac.uk/software/gff/) is often chosen for storing the resulting data [2]. GFF format reports the genomic features in single-line records where information such as type and position are provided. Several tools have already been developed in order to deal with GFF files (refer also to the above URL). Programs such as BEDtools [3], readseq [4] and gff-ex (http://bioinfo.icgeb.res.in/gff/) can perform this task although their usage requires previous command line interface experience. Galaxy [5–7] is an easy to use alternative featuring a graphical user interface which is available as a stand alone version, as well as a web application. Although extremely versatile these programs have limitation in dealing with annotation data. As an example, the gene regions are not straightforwardly reconstructed and instead introns and exons sequences are generated. Enboss [8], together with its graphical user interface (GUI) version Jemboss [9], can also manage annotation files and convert them in a large number of formats. However all these software lack a downstream quality control of the output data (e.g. presence of anomalous characters within the nucleotide sequences, presence of canonical splicing sites in introns, etc.).

Additionally, dealing with annotation data may represent a non trivial task due to the heterogeneity underlying the way a GFF file can be written. Indeed the association between initial and final positions of a feature and its direction (5’ -> 3’ or viceversa), the presence of several splicing variants for the same gene, the possible occurrence of overlapping features should all be considered. This may require an intense scripting effort which can rarely be made by a user with basic informatics skills.

Here we present gff2sequence an open-source program which allows the extraction of gene features from an annotation file while controlling for several quality filters and maintaining a user friendly graphical environment.

gff2sequence (Figure 1) takes in input a GFF (or GTF) annotation file and the relevant multifasta genome information and generates the nucleotide sequences of many genic and intergenic features (e.g. untranslated regions, coding sequences, proteins, introns, exons, genes, transcripts and down/upstream sequences). While the software was designed to work with gene annotation data it can also be used to extract generic features from any multifasta nucleotide sequence (see documentation for details).

Many parameters can be set in order to filter the output sequences. A general quality control can be used for selecting only sequences with no special characters (e.g. N for incomplete assembly, X for masking, etc.), or exceeding a user defined length. Coding sequences can be tested for the presence of a proper start codon (ATG), for the occurrence of a canonical stop signals (e.g. TGA, TAA or TAG), and for the presence of full codons (e.g. the total number of nucleotides is divisible by 3). Such features are automatically selected when the CDS translation is performed (standard genetic code is used to perform this task). Finally introns may be filtered by specifying the splicing signals.

gff2sequence generates three output files when the intergenic sequence information are gathered: (a) a multifasta formatted file reporting the nucleotide sequences, (b) a list of gene couples that are adjacent (e.g. with no intergenic sequences between them) and (c) a list of gene couples that are overlapping (e.g. genes which partially share a portion of DNA).

The software was tested on gff files that were available from a number of curators and performed smoothly for the following species (see Supplementary file “inputFileTested.pdf” for a complete list of the used URL): Arabidopsis thaliana[10], Vitis vinifera[11], (Phytozome [12]), Solanum lycopersicum (Sol Genomics Network at www. http://solgenomics.net/[13]), Oryza sativa[14] (Rice genome annotation project at http://rice.plantbiology.msu.edu/), Zea mays[15] (http://www.maizesequence.org/).

The main algorithm allows fast computations without being too demanding on hardware. As a way of example, full analysis (e.g. quality filters were all selected) of the large Zea mays genome (around 2500 Mbp) was performed in 11 minutes and required between 2.5 and 3 Gigabytes of memory on an Intel® Core™ i5 CPU M 430 working at 2.27 GHz. The presence of 2920 anomalous coding sequences and 928 overlapping genes emerged from such analysis.

We believe gff2sequence may represent a valuable and easy to use alternative for the generation of a customized sequence dataset from general feature formatted file. Moreover identification of anomalous coding sequences, overlapping genes or non-canonical splicing sites may help in refining the automatic gene predictions.