Notes from the biomass

Refuting phylogenetic relationships, published recently in Biology Direct is a peculiar paper in several ways. Bucknam, Bucher and Bapteste suggest a novel way of assessing phylogenies. Rather than searching the most probable of a species tree given a concatenated alignment, unsupported phylogenies are rejected, leaving possible ones and not inferring more than what the data delivers. The method appears powerful in removing weak hypothesis more strictly than classical methods do in difficult cases.

The work builds on the classical phylogeny approaches using conserved markers such as ribosomal proteins and bootstraps calculations, here using Phyml. The difference lies in the evaluation of the tree by two scores that build on different concepts of impossibility. Given two subsets of species on a branchpoint, they estimates the distance between species based on how incompatible their data is. Classical statistical analyses such as Principal Component Analysis and statistical tests can subsequently be used to assess probabilities of assignments.
The philosophy behind it is to develop falsifiable hypotheses whereas currently used approaches are considered to be inherently positivistic. Consequently, the paper starts with lengthy philosophical implications, before introducing the concepts and analysis strategies. The details on operating systems, command line switches and file formats however read like a bachelor thesis from a couple of years back and are in stark contrast to the initial philosophical considerations. Finally, examples show that the methods can be applied to real world problems such as the position of Nanoarchaeum equitans within the archaeal realm (see Figure 2 from the paper below, axis are determined by PCA build on impossibility scores).
The comments of the reviewers greatly enhanced my understanding of the work. Beyond all concerns of publishing policies, the open peer review approach pioneered by Biology Direct also helps the readers to reflect on critical issues of the manuscript. The work is certainly controversial and requires additional research and discussion to tell whether this smart concept is just undersold or but a philosophical consideration with limited potential for application.