FlyBase: introduction of the Drosophila melanogaster Release 6 reference genome assembly and large-scale migration of genome annotations

Description of BDGP Release 6 genome assembly

A reference genome assembly for D. melanogaster was first released in 2000 (8,9). This assembly was based on nuclear DNA sequences derived from the ‘iso-1’ reference strain: isogenic yellow (y¹); cinnabar (cn¹), brown (bw¹), speck (sp¹) (10). Since then, this iso-1-derived reference nuclear genome assembly has been revised to close gaps, improve sequence quality and add sequence from centric heterochromatin regions (2,11).

The latest BDGP Release 6 of the D. melanogaster genome assembly, still based on the iso-1 reference strain, includes a total of 143.7 Mb on 1870 scaffolds comprising 2442 contigs (Table 1). The vast majority, 137.6 Mb, of this sequence resides on seven chromosome arms: X, 2L, 2R, 3L, 3R, 4 and Y (Table 2). Additionally, there are 1862 minor scaffolds that lack precise localization. Almost half of these ‘unlocalized’ scaffolds (n = 884) have been mapped to a chromosome region by comparative genome hybridization in embryos with various chromosome deficiencies: 2CEN (centromere-proximal region of chromosome 2), 3CEN, X, Y, XY (regions mapping to both X and Y chromosomes) and rDNA (ribosomal DNA) (12). The Release 6 reference genome assembly also replaces the previous mitochondrial reference genome assembly, a composite of sequences from various D. melanogaster strains, with one derived exclusively from the iso-1 reference strain (Table 1).

Major improvements in the Release 6 genome assembly compared to Release 5

Release 6 has a number of important changes from Release 5 (Tables 1 and 2). Release 6 is 4.2 Mb larger, even as the total assembly gap length has been reduced to 1.2 Mb, a decrease of 1.5 Mb. The main areas of improvement are to the centric heterochromatin regions of the major chromosome arm scaffolds (X, 2L, 2R, 3L, 3R, 4), which have incorporated over 10 Mb of sequence from minor Release 5 scaffolds (XHet, 2LHet, 2RHet, 3LHet, 3RHet and U). The chromosome Y scaffold has been improved dramatically, increasing over 10-fold in size to 3.1 Mb. Almost all remaining gaps in the chromosome arm scaffolds are in the heterochromatic regions. Small, unmapped scaffolds are now represented individually, instead of being concatenated into pseudoscaffolds (e.g. Release 5 pseudoscaffolds U, 3LHet).

Access to the Release 6 and Release 5 genome assemblies

The BDGP Release 5 and Release 6 genome assemblies have been deposited at NCBI. Full reports are available at NCBI (4), which provides global statistics and quality metrics for each assembly, as well as a ‘full sequence report’ and links to GenBank (13) and RefSeq (14) accessions for individual scaffolds (Table 3).

Beginning with FlyBase update FB2014_04 (released July 25, 2014), all FlyBase data are now reported with respect to the Release 6 genome assembly. However, Release 5 versions of FlyBase, including the last FlyBase update before Release 6 (FB2014_03, R5.57), will remain accessible from the ‘Archives’ menu in the blue navigation bar at the top of most FlyBase pages (‘Other Archives’ sub-menu, http://flybase.org/static_pages/downloads/archivedata3.html). Thus, through these Archives, all Release 5 sequence locations described in gene and genomic feature reports will remain available, as will tools such as BLAST (15) and GBrowse (16) using Release 5. All versions of FlyBase on the Release 5 genome assembly are also available for download on the FlyBase FTP site, available from the ‘Files’ menu of the FlyBase navigation bar (ftp://ftp.flybase.net/releases/).

Feature migration

FlyBase represents as many features located on the genome as possible. Primary among these are the gene model annotations, represented by localized exon features that are joined into one or more models representing transcript isoforms. In addition to the gene models, FlyBase contains numerous other types of localized features including ones that are part of the sequenced genome (e.g. protein-binding sites, enhancers), reagents and variations reported in the literature (e.g. aberrations, mutant alleles, mobile element insertion sites) and discretely aligned features (cDNAs, protein similarities, gene model predictions, RNA-Seq splice junctions). The replacement of the Release 5 genome assembly with Release 6 required migration of millions of discrete features and over a hundred RNA-Seq datasets from the old to the new assembly.

Assembly-to-assembly alignments between Release 5 and Release 6 were generated by NCBI (4), using previously described methods (http://www.ncbi.nlm.nih.gov/genome/tools/remap/docs/alignments), and provided to FlyBase. Using the information provided in these alignments, a map was generated that identified coordinate spans in Release 5 that correspond to equivalent unchanged regions in the new Release 6 assembly. Because the Release 6 assembly contains some regions that are inverted relative to their orientation in Release 5, and other regions that have moved to different scaffolds, the mapping file and necessary coordinate transformations needed to take these types of changes into account. The mapping file underlies the coordinates converter tool (see below) and is available at FlyBase (http://flybase.org/reports/FBrf0225389.html). This map was used to perform the coordinate conversions necessary to update the location of the features from their old Release 5 coordinates and localize them to the new scaffolds of the Release 6 assembly.

Realignment of high-throughput datasets

For many types of high-throughput datasets, direct mapping of reads to the new genome assembly is preferable to the migration process described above, since it allows alignment to newly sequenced regions. However, direct mapping is not always available from the original data producers. Therefore, our policies are as follows.

1. When available, we will replace any FlyBase-migrated dataset with a new analysis directly mapped to the Release 6 genome assembly.

2. For a limited time, we will accept genome datasets mapped to the Release 5 assembly and migrate them to the Release 6 assembly for presentation in FlyBase.

With respect to aligned nucleotide and protein database records presented in FlyBase as evidence, GenBank has provided new Release 6 alignments of cDNA, expressed sequence tag (EST) and proteins using a combination of BLAST (15), ProSplign and Splign (17). GenBank is also producing a Release 6 gene prediction set using their eukaryotic genome annotation pipeline (14).

The modENCODE Transcription Group (http://www.modencode.org/celniker) has also provided FlyBase with de novo alignments of modENCODE RNA-Seq coverage data to the Release 6 genome, including the original developmental profile (18) and the most recent profiles for tissues, treatments and cell lines (19). The new Release 6-aligned datasets will be available on FlyBase by the public update FB2014_06. In the interim, these and other RNA-Seq coverage datasets have been migrated to the Release 6 genome assembly based on the same mapping file used for feature migration (see above). RNA-Seq coverage datasets are displayed in GBrowse and are used to calculate RPKM gene expression values (20). These RPKM values are presented in the High-Throughput Expression sections of gene reports and can be queried extensively using the RNA-Seq Search tool (http://flybase.org/static_pages/rna-seq/rna-seq_search.html). A more detailed description of the FlyBase RPKM calculation is available (http://flybase.org/reports/FBrf0221009.html).

Summary of changes to gene models and other features

The large majority of data were migrated to the new assembly. Unsurprisingly, most of the features that failed to automatically migrate were originally mapped to unlocalized scaffolds (pseudoscaffold U) or to heterochromatic arms that were represented as separate entities in Release 5 (e.g. 2RHet, 3RHet).

Considering gene model annotations, there were 42 gene models that did not automatically migrate to the new assembly because of alterations to the sequence that fell within the gene span. These annotations were manually reviewed and 12 of them were re-annotated on the new assembly. The 30 remaining annotations that failed to migrate had all been located to unlocalized scaffolds in Release 5; 18 of these were ribosomal RNA gene fragments, 11 were Suppressor of Stellate (Su(Ste)) multicopy non-coding RNAs and one was a pseudogene. Additionally, there were 35 annotations that were automatically migrated, but for which the coordinate conversion to Release 6 suggested changes within the annotation. After manual review, 21 of these automatically migrated annotations were accepted without additional changes, 13 were modified to reflect a new CDS and one was deleted (a ribosomal RNA gene fragment).

The improvement of the assembly in heterochromatic regions also led to the resolution of 11 previously fragmented gene annotations. These improvements resulted from relocation of sequence from the Release 5 pseudoscaffold U and unordered ‘Het’ scaffolds (e.g. 3RHet) to the Release 6 chromosome arm scaffolds: Y (six cases), 3R (two cases), 2R, 3L and 4 (one case each). The annotation of these gene models as complete, intact genes is part of the D. melanogaster annotation version R6.02 (FB2014_05), and documented in FlyBase's Release 6 Sequence Assembly Notes, available from the ‘Documents’ menu of the FlyBase navigation bar (http://flybase.org/static_pages/docs/release_notes.html).

Other types of annotated features that did not migrate to the new assembly included 25 annotated mobile elements, seven origins of replication and 54 transcription factor binding sites. In addition, in regions with significant changes, there were aligned and analysis features like ESTs, protein similarities, RNA-Seq alignments and RNA-Seq-derived exon junctions that failed to migrate. While we migrated as many of these types of analysis features as possible, our ultimate goal is to obtain new alignments and analysis on the Release 6 assembly (see above).

Conversion tools for users

Many researchers will want to convert the Release 5-based coordinates for their own data into the corresponding Release 6 coordinates. For small sets of data, FlyBase provides the Coordinates Converter tool, which converts genomic coordinates based on a specific genome assembly release to those for a later release. For each set of coordinates converted, the tool also reports if there have been any changes to the genomic sequence in that region between the assembly releases. For example, for the input Release 5 coordinates ‘2LHet:16,945..17,066’, this tool returns the corresponding Release 6 coordinates ‘2L:23,158,471..23,158,593’, with the notes ‘includes 1 area of change; different scaffold.’ The Coordinates Converter may be accessed from the ‘Tools’ menu on the navigation bar, and offers a number of input and output options (Figure 1). A separate Release 6-to-Release 5 backwards converter is also provided (Figure 1).

While the FlyBase Coordinates Converter tool can be used for lists of genomic coordinates, the NCBI Genome Remapping Service is recommended for the conversion of large datasets (www.ncbi.nlm.nih.gov/genome/tools/remap). For D. melanogaster, users of this remapping service can upload files and convert coordinates between Releases 5 and 6 (in either direction). A variety of file formats are supported for input as well as output. In addition to the main ‘Annotation Data’ file, summary and mapping reports are provided.

Download of the Release 6 genome assembly sequence

Sequence data for the D. melanogaster genome assembly can be downloaded from the FlyBase FTP site, accessible from the ‘Files’ menu in the FlyBase navigation bar (ftp://ftp.flybase.net/). Various cuts of the data are provided in multiple formats, ranging from GFF3 (http://www.sequenceontology.org/gff3.shtml), FASTA (21) and GTF (http://mblab.wustl.edu/GTF2.html) files to Chado XML (22) data sliced into portions such as scaffolds, gene models, etc. These data are provided for the current genome release and all older releases. The latest, most up-to-date D. melanogaster genome can be found in the /genomes/Drosophila_melanogaster/current subdirectory; within that location, raw scaffold data are in the /dna directory (e.g. dmel-raw_scaffolds-r6.xx.tar.gz). The FlyBase FTP site requires an anonymous login in passive mode; these are the defaults for most FTP clients and browsers, but if you experience any difficulties please check your settings.

BLAST

FlyBase provides a stand-alone BLAST (15) server for 50 different arthropod genomes (http://flybase.org/blast/). With each FlyBase release, the BLAST tool is updated with the latest genome and annotation data. Thus, as of FB2014_04, Release 6 data have been underlying BLAST alignments for D. melanogaster. Archived versions of FlyBase may be used to perform BLAST alignments against Release 5 genomic sequence and gene model annotation sets. FlyBase BLAST offers standard BLAST options (such as TBLASTN) as well as advanced options where users can specify the matrix, word size, codon bias and complexity filters. When using the ‘Genome Assembly’ database option, GBrowse buttons will appear in the results next to each BLAST hit, allowing users to directly map their hits onto the D. melanogaster genome. This feature is also available for the other 11 Drosophilid genomes that are represented in FlyBase GBrowse. This allows users to see the BLAST hit within the context of the surrounding genome, which can be particularly useful when assessing whether a hit is biologically significant or whether to study a candidate gene further.

GBrowse 2

For the past decade, FlyBase has used the generic genome browser, GBrowse (16), for the display of genome annotations and genome-aligned evidence on the reference genome assembly. GBrowse allows users to navigate to a region of interest using coordinates or a landmark (e.g. a gene name or feature identifier), and from that entry point, zoom or scroll along the genome. The feature types displayed are user-selected. Genomic sequence for the region in view can be downloaded in a variety of formats. Concurrent with the introduction of Release 6, FlyBase has also upgraded to GBrowse 2 (23). GBrowse 2 retains the same functionality described above for GBrowse. However, GBrowse 2 contains some important upgrades, particularly its ability to handle much more data, which has allowed FlyBase to offer all D. melanogaster data options on a single view. GBrowse 2 also offers users easier track customization (vertical placement, open or hidden views) and limited smooth-track panning (side-to-side sliding). A new lasso feature can be used to select a region of interest, and either navigate to that region or download the associated genomic sequence (Figure 2). GBrowse 2 is accessible from the homepage and various FlyBase report pages.

Recommendations for referencing public D. melanogaster genome data

With the replacement of the reference genome assembly, as well as the frequent changes to annotation sets, it is important that researchers, during publication and when submitting to public data repositories, indicate clearly the versions of genome assembly and FlyBase annotation datasets that were employed in a particular analysis. Genome assemblies can be unambiguously identified by the corresponding NCBI accession (Table 3). FlyBase annotation sets can be unambiguously identified by the FlyBase release number or the gene model annotation set version number. For example, the September 2014 public update of FlyBase (FB2014_05) includes ‘release notes’ that specify Release 6 as the reference genome assembly, and version R6.02 as the FlyBase D. melanogaster gene model annotation set for this update. Citing either FB2014_05 or FlyBase R6.02 is sufficient to uniquely point to this public dataset. The release notes are accessible through the ‘Documents’ menu in the FlyBase navigation bar (http://flybase.org/static_pages/docs/release_notes.html). All FlyBase pages indicate the FlyBase version in the header; in GBrowse 2, the header also indicates the FlyBase annotation version in view (e.g. R6.02). Archived release notes for previous FlyBase updates (and the release dates) are available through the navigation bar's ‘Archives’ menu (‘Other Archives’ sub-menu), so that researchers can go back to identify the relevant FlyBase version from which data were obtained on a given date.

While FlyBase updates gene model annotations and other molecular features with each bimonthly public database update, some users prefer to update less frequently. For such users, FlyBase recommends using the frozen datasets that we submit annually to GenBank and RefSeq at NCBI; the current FlyBase submission to NCBI corresponds to FlyBase D. melanogaster annotation set version R6.01.

Adoption of the Release 6 reference genome assembly by the Drosophila research community

Uniform adoption of the new Release 6 reference genome assembly by researchers and the resources they use will be a gradual process. It is our understanding that the Release 6 assembly will have replaced Release 5 at the UCSC genome browser (7) and Ensembl (24) by January 2015 or earlier. Time lines for similar updates at other public D. melanogaster resources are not known to us. Additionally, many ongoing research projects are surely based on the Release 5 reference genome assembly. We advise researchers to be mindful of this issue when using a publicly available dataset or resource, and to identify the relevant reference genome assembly and FlyBase gene model annotation set version used for a particular analysis.