Publisher's Award for Excellence in Systematic Research

Julia and collaborators assess the validity of alignment-free based methods such as k-mer based distances for phylogenetic classification and compare them with traditional multi-gene alignments. They demonstrate the accuracy of k-mer based distances when applied to highly incomplete genomes, showcasing their potential utility in characterizing cryptic species from environmental genome data. This type of data is common in unicellular algal and bacterial systematics. This method raised great interest among scientists working in the field of environmental genomics and meta genomics.

Lenard and collaborators investigate the link between long-branch attraction and across-site compositional heterogeneity in deep phylogenetics via simulations and empirical datasets. They also propose a new pipeline based on the CAT model (CAT-PMSF) to assess the risk of compositional heterogeneity bias when the CAT model does not converge. Given the computational effort required to run a CAT analysis on a large genomic dataset, this work is particularly relevant. The authors do not only offer a tool to help researchers be aware of the effects of this model deviation on phylogenetic accuracy, but also a way to solve it.

Jeremy and collaborators introduce a new model to compute the joint probability density of a phylogenetic tree and a record of occurrences: e..g, taxa with no associated data such as fossils representing extinct lineages with known stratigraphic age, or epidemiological patients for which there is no associated sequence. It extends a previous approach by the senior author, adding a piece-wise process that allows discrete shifts in the rate of parameters such as speciation, extinction and sampling. One interesting aspect of the model is that it can be used with mixed datasets, where some tips are associated with data (fossils with coded morphological characters) and others are not (fossil records from the Paleobiology Database). They implement their Occurrence Birth-Death Process (OBDP) as a new distribution in the Bayesian software RevBayes, allowing it to be used as a tree prior for morphological state reconstruction or biogeography.

Yueyu and collaborators tackle a common problem in current phylogenomics, placing a sequence/s obtained from a single gene into a phylogenomic backbone. The common approach is to assume the that new sequences and the genomic reference phylogeny evolved under the same molecular evolutionary process. Here, they present a deep learning algorithm that learns to extend a genomic species tree without any prior knowledge of the model, and with levels of accuracy similar to model-based approaches. The method is very timely, especially for the fields of epidemiology and environmental DNA, and in line with recent advances in the field of online phylogenetics.