FlyBase Comparative Genomics Data

Jump to: navigation, search

Comparative Genomics Data Implementation in Chado

by Peili Zhang, David Emmert, Pinglei Zhou, William Gelbart, and the FlyBase Consortium


The data implementation standard of comparative genomics in Chado is described in this document. The stated data implementation has been formulated by the developers at FlyBase ( and adopted by FlyBase for production. For the purpose of clarity, pairwise comparative genomics results are being employed in this document as examples to illustrate the points. Yet the same implementation can be readily generalized and applied to multiple-genome comparative results. To store the comparative results between two genomes in Chado, the sequence and annotation data of each of the relevant genomes need to be implemented independently in the same Chado database instance. The Chado standards of genomic sequences and annotation are detailed in separate documents, which should be followed before attempting the comparative data implementation described here. Comparative genomics data of interest are grouped into the following two categories: orthology and synteny, for which the Chado standards are illustrated separately below.


a) orthologous regions

Figure 1:implementation of orthologous region features in Chado. Two Drosophila genomes are used as examples. Dmel: Drosophila melanogaster; Dpse: Drosophila pseudoobscura.

Orthologous regions are implemented similarly to alignment match features, the only difference being that the orthologous regions are features of type 'orthologous_region' whereas alignment match features are of type 'match'. Basically, the orthologous regions are stored as analysis features of type 'orthologous_region', belonging to the fake organism of genus 'Computational' and species 'result'. They are linked to the appropriate analysis that derived them through analysisfeature links. Each orthologous region derived from TBLASTN analysis, as an example, has three featureloc records of ranks 0 to 2, with rank = 0 for the localization on genome 1, rank = 1 on the protein feature of genome 1, and rank = 2 on genome 2, for a pairwise comparative genome analysis between organism 1 and 2. The implementation of orthologous regions is illustrated in Figure 1 below.

b) orthologous genes

Figure 2: Chado representation of orthology relationship between a gene pair.

The orthology relationship between a gene pair is represented in Chado through feature_relationship, with the two genes being object and subject (arbitrarily set and consistent across the database) and the feature_relationship type being 'putative_ortholog_of'. The detailed data storage is depicted in Figure 2.


Syntenic regions between genomes are stored as features of type 'syntenic_region', belonging to the made-up organism of genus 'Computational' and species 'result'. The genomic spans of these syntenic regions on each of the genomes are stored as featureloc records, with featureloc.rank assigned differently for the different genomes. For example, genomic span 10000..50000 of genome 1 chromosome 1 is syntenic to the span 12345..56789 of genome 2 chromosome 2, a feature of uniquename 'syntenic_block:1', type 'syntenic_region' and organism -'Computational result' is created. The feature 'syntenic_block:1' has two featureloc records: 1. rank = 0, fmin = 9999, fmax = 50000 (note: the interbase coordinates are utilized), strand = NULL, and srcfeature_id pointing to genome 1 chromosome 1; 2. rank = 1, fmin = 12344, fmax = 56789, strand = NULL, and srcfeature_id pointing to genome 2 chromosome 2.

In addition, the component orthologous regions of a syntenic region are recorded through feature_relationship records between the syntenic region feature and the relevant orthologous region features which are implemented as described in Section I. a), with a feature_relationship type of 'partof' and feature_relationship.rank left as the default value of 0. Bear in mind that the orthologous region features are, in essence, 'match' features, each of which represents a pair of orthologous genomic spans from the two genomes. The reason the feature_relationship.rank field is left as default 0 is that the relative ordering of the orthologous regions on each genome is implied by their genomic locations. Furthermore, one pair of orthologous regions may be in slightly different relative ordering on the two genomes, for example, orthologous regions 1, 2, 3, 4 form one syntenic block. Their genomic spans on genome 1 are in the order of 1, 2, 3 and 4, while their spans on genome 2 are in the order of 1, 3, 2, and 4 where the relative ordering between the 2nd and 3rd orthologous regions is reversed as compared to genome 1. Therefore, it is not always possible to assign one set of ranks for the orthologous regions in a syntenic block to truthfully reflect the ordering of these regions in each individual genomes.

The implementation of the syntenic relationship between genomes is schematically shown in Figure 3.

Figure 3: Chado representation of syntenic relationship between a pair of genomes, based on orthologous_regions shown in Figure 1. Dmel: Drosophila melanogaster; Dpse: Drosophila pseudoobscura.