FLIP: Future Likelihood Inferences for Phylogenetics

In times of increasingly efficient sequencing technologies bioinformatics is facing new challenges to handle and analyze the previously unknown flood of data on molecular sequences and analyze. Parallel to the revolution in molecular biology the computer technology has made substantial progress. In the proposed project we want to use bioinformatics, statistical and computer science methods to work efficiently with the new data and new technologies. Our main focus will be placed on the reconstruction of phylogenetic relationships of contemporary organisms based on their genes. In a first step, we will further improve known methods of phylogenetic reconstruction by utilizing the progress computer technology. We want to use the technological development, to save computing time and at the same time also process more data per unit of time, which not only saves the researchers time but also reduces the electricity consumption. In addition to improving existing methods, we will also develop completely novel statistical estimation procedures and algorithm to cope with the flood of data. These methods will be integrated into computer programs that will be made available to the scientific community. In this way, we want to get feedback, which should lead to further improvement and flexibility of our programs. Our method will allow the reconstruction of phylogenetic relationships for very many living organisms. These relationships are calculated on a solid statistical basis and allow conclusions about the plausibility of the phylogeny. Thus, ultimately our project will bring biology closer to its goal to calculate the tree of life.

With the ever-growing biological data the efficiency and accuracy of computational methods for phylogenetic inference has become an important question towards reconstructing the tree of life. The constant improvement of the software and algorithms allows for the analyses utilizing complex models of evolution on massive datasets. Moreover, it is extremely important to take into account the intrinsic properties of the data for the improvement of the algorithms. During the funding period we have attained a number of goals. First, we provided a parallelised version of our flagship software for phylogenetic inference IQ-TREE. We also introduced the method to improve parameter estimation under one of the most used evolutionary model, namely the invariable sites plus model. We released a second version of UFBoot method, frequently used to assess the confidence of evolutionary trees in maximum likelihood inferences. Moreover, we introduced a similar method MPBoot for parsimony inference. We also provided a tool for fast and accurate evolutionary model selection ModelFinder, where we introduced a distribution free method to assign evolutionary rates to individual sites in a DNA or a protein. Among other developments is the first-ever production level implementation of tree inference accounting for the so-called phylogenetic terraces, which occur in the presence of missing data. Overall, the project was very successful and resulted in many theoretical developments, which were implemented in IQ-TREE and can be freely used by practitioners for the inference of evolutionary relationships. (The software is available at www.iqtree.org).