Existence of non-unique neighbor-joining phylogenetic trees of microsatellite markers

Abstract

Abstract Background: Neighbor-joining (NJ) is a distance-based method used to analyse and visualise data that generates a phylogenetic tree. It is found in a wide range of scientific research and is upstream of all areas of bioinformatics and computational biology. NJ's drawback is the generation of different binary phylogenetic trees if more than two clusters of elements have the same distance in any step of the phylogenetic tree generation process. We analyse the extent of this problem in a particular context: the clustering of microsatellite markers with NJ. For this purpose, we calculate the proportion of publications in the Scopus database in which more than one binary phylogenetic tree is possible. Results: We have observed that 13% of articles can produce more than one binary phylogenetic tree; thus, results shown in these articles would have to be deeply analysed or their conclusions revised. To illustrate the problem from a practical point of view, we have selected an example where more than one binary phylogenetic tree solution is possible. Conclusion: Articles that show only one phylogenetic tree, although NJ could generate several of them, are somewhat biased and hard to reproduce. The percentage of publications with non-unique NJ phylogenetic trees is large enough to be considered a general problem in our scientific community.