This file describes the steps used to create the salmonid phylogenies data set, and
a description of the contents of the archive file found on the cGrasp website.


Contigs

  EST sequences were taken from a variety of sources (Koop lab, Genome Sciences
Centre, NCBI) for 6 species: Salmo salar (Atlantic Salmon), Oncorhynchus mykiss
(Rainbow Trout), Osmerus mordax (Rainbow Smelt), Coregonus clupeaformis (Lake
Whitefish), and Salvelinus fontinalis (Brook Trout). For each species a two-stage
assembly was performed using Phrap, with values of 100/0.99 and 300/0.96 used
for Phrap's minscore and repeat_stringency parameters for the two stages. The
output of this step was a set of contigs for each species as summarized by the
following table.

Species # of ESTs # of Contigs
Atlantic Salmon (ssal) 434384 81398
Rainbow Trout (omyk) 346704 51199
Rainbow Smelt (omor) 36758 12159
Lake Whitefish (cclu) 10842 6446
Brook Trout (sfon) 10051 4946
Artic Grayling (tthy) 5930 5408


Clusters

  All of the contigs resulting from the two-stage assemblies in step 1 were then
blasted against each other (evalue=1e-35, hits==100) and the results used to
generate clusters of contigs. Clustering algorithm:

for each contig c
	for each blast hit against a different contig
		if alignment good
			a = existing cluster containing contig or new cluster if none
			b = existing cluster containing blast hit or new cluster if none
			join clusters a & b

where alignment consisted of ends-free alignment with scores of 2/-2/-5/-1 for
match/mismatch/open gaps/extend gaps. A good alignment was considered to be
greater than 75% (of the length of the shorter sequence) overlap with greater than
70% identity in the overlapping region.


Trimming the Clusters

  After the contigs had been grouped into clusters the individual clusters were then
further filtered to only contain contigs that had mutually overlapping regions. Cluster
filtering algorithm:

for each cluster
	sort contigs in cluster by length (longest to shortest)
	good set = {}
	for each contig in cluster
		if alignment good with all other members of good set
			add contig to good set
			trim all members of good set to largest common alignment
	cluster = good set

where alignment consisted of ends-free alignment with scores of 2/-2/-8/-1 for
match/mismatch/open gaps/extend gaps. A good alignment was considered to
be greater than or equal to 300 base pairs overlap with greater than 60% identity
in the overlapping region.


Removing Gaps

  Prior to making distance matrices all gaps (and their corresponding positions in
other sequences of the cluster) were removed. After this point all clusters consisted
of one or more sequences of exactly the same length and with no gaps. Clusters not
containing at least one representative from all 6 species were discarded, which
resulted in 78 clusters remaining.


Bootstraps, Distance Matrices and Consensus Trees

  Each cluster was bootstrapped 500 times, and then distance matrices were computed
for each cluster using the F84 model of nucleotide substitution and Gamma-distributed
rates of variation across sites with a coefficient of varation of 0.5. A neighbor-joining
tree was then computed from each distance matrix, and a 70% majority consensus tree
was constructed for each resulting data set. The trees were then rooted manually using
RainBow Smelt as an outgroup. As a final step pairs of leaf nodes of the same species
were iteratively collapsed to simplify the trees (since duplication events occuring after all
speciaition events aren't of interest to this study).


Salmoid Phylogenies Data File

  The salmonid phylogenies data file available on the cGrasp website contains a summary
of the resulting data sets. The file is a gzipped tar archive containing one numbered
subdirectory for each of the 78 data sets (the numbers are arbitrary and have no
specific meaning). Each subdirectory contains:

  - The final alignment for the cluster that was used to create the tree, in interleaved
    phylip format.
  - A tab-separated text file (*.csv) containing detailed information about the sequences
    that make up the cluster (Internal Koop lab sequence IDs, Swissprot annotation
    information, and accession numbers for the individual ESTs that make up each
    individual contig)
  - A “raw” tree file, prior to collapsing identical leaf nodes and removing Koop lab
    ID numbers.
  - A “processed” tree file, which has the identical leaf nodes collapsed and the leaf
    node names converted into 4-letter species name abbreviations.