Highlight talk – Theme: Genome
Abstract
It has long been believed that filtering out the unreliable regions of Multiple Sequence Alignments (MSAs) leads to a higher signal-over-noise ratio and in turn better phylogenetic trees. With several automated alignment filtering methods proposed in recent years, there is still lacking a systematic and comprehensive evaluation of different alignment filtering methods in the context of phylogenetic inference from real data. We performed phylogenetic tests of alignment accuracy into the unfiltered and filtered alignments and evaluated the performance of different filtering methods on single-gene tree reconstruction. Contrary to the current beliefs, we found that filtering alignments usually gave worse phylogenetic tree than the unfiltered alignments, on both genome-wide empirical and simulated data sets for a wide range of parameters and methods. This finding suggests using the unfiltered alignments in phylogeny inference and guide the development of better filtering methods.
Authors
Ge Tan, Imperial College London, United Kingdom
Matthieu Muffato, EMBL-European Bioinformatics Institute, United Kingdom
Christian Ledergerber, ETH Zurich, Switzerland
Javier Herrero, University College London, United Kingdom
Nick Goldman, EMBL-European Bioinformatics Institute, United Kingdom
Manuel Gil, Zurich University of Applied Sciences, Switzerland
Christophe Dessimoz, University College London, United Kingdom,