Three types of homology relationships are provided.
Protein trees are calculated using the longest peptide of all the Ensembl protein coding genes. Proteins are clustered based on Best-Reciprocal Hits and Blast Score Ratios. Each cluster of protein is aligned using Muscle. NJTREE/PHYML is used to get a gene tree from each multiple alignment. The gene tree is reconciliated with the species tree to call duplication events.
Homologues (orthologues and paralogues) are deduced from the protein trees. These are built using the longest transcript of each gene only. A detailed description of the method is provided here.
Families are groups of related genes. The full set of Ensembl proteins is extended using all the metazoan proteins from SwissProt and SPTrEMBL.
BlastZ-net (Schwartz S et al., Genome Res.;13(1):103-7, Kent WJ et al., Proc Natl Acad Sci U S A., 2003;100(20):11484-9) alignments are provided for closely related pairs of species. The alignments are the results of post-processing the raw BlastZ results. In the first step, original blocks are chained according to their location in both genomes. The netting process chooses for the reference species the best sub-chain in each region. The reference species in the BlastZ-net alignments is defined in this table:
[[INCLUDE::/info/data/compara/blastz_net.inc]]Translated blat (Kent W, Genome Res., 2002;12(4):656-64) is used to look for homologous regions between more distantly related pairs of species. We expect to find homologies mainly in coding regions.
[[INCLUDE::/info/data/compara/tblat.inc]]Pecan is used to provide global multiple genomic alignments. First, Mercator is used to build a synteny map between the genomes and then Pecan builds alignments in these syntenic regions.
Pecan is a global multiple sequence alignment program that makes practical the probabilistic consistency methodology for significant numbers of sequences of practically arbitrary length. As input it takes a set of sequences and a phylogenetic tree. The parameters and heuristics it employs are highly user configurable, it is written entirely in Java and also requires the installation of Exonerate. Read more about Pecan.
Additionally we use Gerp (Cooper GM et al., Genome Res., 2005; 15:901-913) to calculate conservation scores and call constrained elements on the 10 way multiple alignments. Conservation scores are estimated on a column-by-column basis. Constrained elements are stretches of the multiple alignment where the sequences are highly conserved according to the previous score.
[[INCLUDE::/info/data/compara/pecan.inc]]We calculate syntenic regions using blastz-net alignments. We look for stretches where the alignment blocks are in synteny. The search is run in two phases. In the first one, syntenic alignments that are closer than 200 kbp are grouped. In the second phase, the groups that are in synteny are linked provided that no more than 2 non-syntenic groups are found between them and they are less than 3Mbp apart.
[[INCLUDE::/info/data/compara/synteny.inc]]