Quicktree - SD Stephan

Phylogenetic methods are becoming part of the standard methodologies used in the analyses of biological sequence data, but several of the classical tools of phylogenetic inference were not designed with high throughput applications in mind. During our surveys, we haven't found software for the fast inference of neighbor joining trees from sequence alignments which could use biologically realistic substitution models with protein alignments. Thus, we developed quicktree-sd based on an e cient implementation of the neighbor joining algorithm by [3], by implementing an amino acid distance correction based on Scoredist distances. The tool is available at ftp.awi.de. High throughput phylogenetic inference is being used to an increasing extent in the context of genomic data. On the one hand, phylogenetic methods are used to improve genome annotation [1],[2]; on the other, genomic data are used in order to understand evolutionary / phylogenetic questions [4]. One of the simpler methods of phylogenetic inference is neighbor joining (NJ; [5]), which provides a good compromise between computational requirements and accuracy, provided that biologically realistic substitution models are used for estimating distances between sequences [7]. Surprisingly, we have not found any available tool for constructing NJ trees from sequence alignments combining performance with a biologically realistic amino acid substitution model. Quicktree [3] was developed as a tool for the fast inference of NJ trees, with high throughput applications in mind. However, the original implementation of Quicktree did not use an amino acid substitution matrix for calculating the distance matrices. Scoredist was proposed as a simple and generally usable protein sequence distance estimator by [6]. We implemented this distance estimator in quicktree and provide the tool for the scienti c community as source code and binaries at ftp.awi.de (also see www.awi.de/en/go/bioinformatics).