Construction of phylogenetic trees on parallel clusters

In this work, we present the preliminary step of a novel approach for the construction of phylogenetic trees on large parallel clusters of PCs. Computation of multiple alignments of biological sequences and phylogenetic tree construction are performed simultaneously. Any algorithm built upon this process uses the concept of neighborhood (which can be informally defined as sets of evolutionary related sequences). The process, called PhylTre, schematically consists in three iterative steps: the first step produces an undirected graph from a pre-processing operation. The second step aims at determining a neighborhood for each sequence. The third step builds partial phylogenetic trees using results stemmed from step two. The steps are applied iteratively until the whole phylogenetic tree is obtained. A sequential code is available and it is currently implemented in parallel on a large cluster of PCs available at ID-IMAG.