All the genes sequenced in the present study except GH7 from R. miehei that was too short to produce any significant alignment and the genes found in NCBI (see Table 3) were aligned with MUSCLE before curating the alignments with Gblocks and building of phylogenetic trees with PhyML (Dereeper et al. 2008).
First, −35 and −10 elements were extracted from the input promoter sequences, aligned, and separate PSSMs constructed from the alignments with a pseudocount=1.
The performance of the method is comparable with that of the best ones from 10 aligners tested, regarding both the quality of the alignments with respect to hand curated ones, and the classification ability.
In contrast to some previous studies [20], [21], we observed a negative correlation with the number of sequences in these alignments, i.e. the alignments with a larger number of sequences were less well aligned.
Considering that insertions and deletions in sequences strongly affect the position of HSR, we first aligned the orthologs using MUSCLE (Edgar 2004), and then excluded the alignments with insertions and deletions >10 nt.
During the SNP calling, the alignments with low mapping quality (MAPQ score <10) were skipped.
The pre-computed transformation matrices to obtain the alignments with the subset can be effectively recombined to generate new alignments.
The fringe distances and directions result from the different angles in the alignments with no effect to the fringe visibility.
For example, at 0.75 Transform-Tanimoto, 25% of the alignments with a 0.75 ROCS shape Tanimoto do not share an associated AR reference shape.
These sequences must be compared to the TRBD1 and TRBD2 gene segments and the alignments with most identical nucleotides are kept as TRBD candidates.
Model structures from the top five families were obtained from the alignments with the top scoring member of the family.
