For visual validation, we employ the features as input for hierarchical clustering using the dynamic tree cut method that is included in CheS-Mapper to automatically compute the number of clusters [46].
Obviously, this is only a very rough way of evaluating sequence-comparison methods, since the resulting tree topologies not only depend on the distance values calculated by the evaluated methods, but also on the tree-reconstruction method that is applied to them.
Here, we present DACTAL ('Divide-And-Conquer Trees (ALmost) without alignments'), a method that is designed to estimate trees, but not a MSA, on large datasets.
Therefore, the following corollary can be easily proven: Corollary 1: If the starting tree is computed using a method that is statistically consistent under the multi-species coalescent model, then the pipeline based on either the DACTAL or SSG decomposition is statistically consistent under the multi-species coalescent model.
By comparison, TreeFitter is an event-based parsimony method that is useful for exploring the relationships between gene trees (parasite tree) and species trees (host tree).
The supertree approach seeks to amalgamate the information from many input trees and therefore is a sequence-based method that is more democratic in terms of getting its information.
In particular, ASTRAL-2, a species tree estimation method that was initially designed to estimate species trees in the presence of ILS, had excellent accuracy and generally gave somewhat more accurate results than the other methods we explored.
In contrast, the subtree to the left of the root (or the blue cluster defined by K-means clustering) contains only strategies that use species tree construction methods that are not strictly topologically-based (i.e., Concatenation, SMRT, or STEAC).
On the other hand, as pointed out in [ 6], in order to make quartet-based methods work it is extremely important to obtain 4-taxon tree reconstruction methods that are not affected by the presence of long-branch attraction [ 7].
Nevertheless, if we want to employ quantitative methods from biology to supplement our research in historical linguistics, it could be much more fruitful to get away from focusing exclusively on those methods that yield simple family trees, and instead look for methods that were designed to handle lateral transfer.
