A resampling strategy for reliable network construction

In the classical approach to tree reconstruction schemes, such as pair group methods, maximum parsimony or minimum spanning trees, two major problems are not addressed at a fundamental level. First, for numerous kinds of experimental data, these methods produce equivalent solutions, but provide no way of handling those degeneracies. Second, the real-life data fed to these methods is treated as exact data, and possible measurement errors cannot be taken into account. We provide a statistical solution for both the degeneracy and data imperfection problem, which is built as a framework around the clustering method. It is therefore independent of the particular choice of clustering or population modeling algorithm and is applicable to any of the presently known methods that are subject to one or both of these problems.