Skip to main content
Figure 2 | BMC Bioinformatics

Figure 2

From: FIGENIX: Intelligent automation of genomic annotation: expertise integration in a new software platform

Figure 2

Phylogenomic inference pipeline. For more details about all the steps and functionalities automated in the pipeline see material and methods sections of the 2002 and 2003 phylogenomic papers [29, 30]. From the query sequence, a dataset of putative homologous sequences is first built by BLAST [16] on a protein database like NR. We filter raw dataset to eliminate sequences potentially non-homologous, disturbing alignments and doubles. User can choose to focus on a specific scope on any node of the tree of life (the vertebrates, the bilaterians...). In the next step, we produce an alignment with CLUSTALW [19]. Then the alignment is modified to eliminate large gaps. Since phylogenetic analysis is done at the domain level, we next detect these domains with HMMPFAM [23]. For each domain alignment (extracted from the original alignment), a bias correction phase is run, to eliminate: – Non-monophyletic "repeats" in a tree built with NJ [31] algorithm on CLUSTALW software. – Sequences with a diverging composition by using an amino-acid composition test of TREE-PUZZLE software [22] (with an alpha risk set to 5%). – Sites not under neutral evolution [35]. Once domains are "purified", and after congruent domains selection with HOMPART test from PAUP package [20], a new alignment is built by merging preserved parts of domains' alignments. From this alignment, three phylogenetic trees are generated using NJ, ML (with TREE-PUZZLE [22]) and MP (with PAUP [20] package) methods. By comparing topologies of these trees with PSCORE command ("Templeton winning sites" test) from PAUP package and KISHINO-HASEGAWA [34] test from TREE-PUZZLE package, fusion of these trees in a unique consensus tree is produced. Through the comparison of this consensus protein tree with a reference species tree, (the tree of life from NCBI [26]), we then deduce orthologous proteins to the query sequence.

Back to article page