([frækkert] danish for cheeky person, also affectionate)

The goal is to provide a method that allows to draw phylogenetic trees from allele frequencies of few but hyperpolymorphic genes, while doing reliable bootstrapping. For example genes belonging to the MHC have many variable loci, leading to many different alleles. Conventional methods, however, are not suitable for low numbers of marker genes (Phylip, DISPAN, check GenPop), since they perform bootstrapping with gene loci as units.

E.g., in https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0169929 the authors claim that the provided phylogeny (in Figure 2) has 100% bootstrap values. This tree however is just based on a single gene (HLA-DRB1), and the used tool (DISPAN) "resamples" exactly from that one gene only, thus inevitably producing identical bootstraps. In other words, if only allele frequencies from a single gene are used, conventional bootstrapping performs resampling with replacement, draws repeatedly from only a single locus. Trivially, this single locus is in agreement with itself, and the tools report 100% branch support for all clades in the tree, which leads to false confidence.

Sample usage:

look at comprehensiveTree.py

Python >=3.6, Various python libraries

• numpy
• scipy (for Nei distance calculation)
• BioPython
• pandas
• dendropy
• networkx

Maybe the latter two can be skipped.