Browse/search for people

Publication - Professor Davide Pisani

    Inadvertent Paralog Inclusion Drives Artifactual Topologies and Timetree Estimates in Phylogenomics


    Siu-Ting, K, Torres-Sánchez, M, Mauro, DS, Wilcockson, D, Wilkinson, M, Pisani, D, O'Connell, MJ & Creevey, CJ, 2019, ‘Inadvertent Paralog Inclusion Drives Artifactual Topologies and Timetree Estimates in Phylogenomics’. Molecular Biology and Evolution, vol 36., pp. 1344-1356


    Increasingly, large phylogenomic data sets include transcriptomic data from nonmodel organisms. This not only has allowed controversial and unexplored evolutionary relationships in the tree of life to be addressed but also increases the risk of inadvertent inclusion of paralogs in the analysis. Although this may be expected to result in decreased phylogenetic support, it is not clear if it could also drive highly supported artifactual relationships. Many groups, including the hyperdiverse Lissamphibia, are especially susceptible to these issues due to ancient gene duplication events and small numbers of sequenced genomes and because transcriptomes are increasingly applied to resolve historically conflicting taxonomic hypotheses. We tested the potential impact of paralog inclusion on the topologies and timetree estimates of the Lissamphibia using published and de novo sequencing data including 18 amphibian species, from which 2,656 single-copy gene families were identified. A novel paralog filtering approach resulted in four differently curated data sets, which were used for phylogenetic reconstructions using Bayesian inference, maximum likelihood, and quartet-based supertrees. We found that paralogs drive strongly supported conflicting hypotheses within the Lissamphibia (Batrachia and Procera) and older divergence time estimates even within groups where no variation in topology was observed. All investigated methods, except Bayesian inference with the CAT-GTR model, were found to be sensitive to paralogs, but with filtering convergence to the same answer (Batrachia) was observed. This is the first large-scale study to address the impact of orthology selection using transcriptomic data and emphasizes the importance of quality over quantity particularly for understanding relationships of poorly sampled taxa.

    Full details in the University publications repository