Principal Investigator

Scott   Keogh BOZO, Science, ANU

Project Title

Direct Optimization and Long Branch Attraction

Brief Description for General Publications

The project will assess whether direct optimization (DO), a novel phylogenetic method, exhibits greater systematic bias towards incorrect hypotheses than standard maximum parsimony methods. We will address this question by creating a large number of simulated data sets on known phylogenies under a variety of conditions and analyzing them under both direct optimization and maximum parsimony to assess accuracy of reconstruction. We intend to perform the data simulation and maximum parsimony analyses on desktop computers in our own lab, but need the automation and computing power of APAC to perform the DO analyses. For each set of conditions we will analyze 100 multiple sequence alignments each consisting of 1000 nucleotide characters simulated for four taxa. Each alignment will be analysed using DO under ten alignment parameter sets. This will result in 1000 analyses per condition set. For this initial study we intend to analyze twenty condition sets, resulting in 20000 DO analyses. Initial experiments on desktop PCs have indicated that each analysis takes between six and ten seconds, resulting in total computing time of 33-55 hours. DO is implemented in the software OY (and a parallelised version POY) that is freely available for download from the internet. One of us (M. Morgan) has held a preliminary meeting with Roger Brown at APAC to assess the feasibility of this study. Roger has informed us that the linux version of OY canbe successfully installed and run at the facility with little to no problems. If this trial is successful we will extend our simulation study to six other models with four taxa (6x 50 SU) and also to models with >4 taxa (potentially a few thousand SU).