Reconstructing the history of a sample of nuclear sequences must take into account the possibility of reticulate evolution caused by recombination or gene conversion. Many methods have been proposed to divide the alignment of sequences subject to recombination into regions with different phylogenetic histories. The history of the sample could then be recovered by `glueing' the histories of different regions together. Here, four different methods to detect recombinations are chosen and investigated with respect to their ability to detect recombination events in simulated samples of DNA sequences. The coalescent process with recombination and mutation (Jukes-Cantor model) is used to generate samples of sequences subject to recombination and sample histories with recombination events. Because the applicability of the coalescent process is well-understood in a population genetic context, it provides a reasonable framework for simulations. The methods are evaluated in different set-ups.
First, we have simulated samples which have experienced exactly one recombination event in their history. We applied the methods to these samples, and registered how often the recombination could be retrieved.
Second, we have simulated samples without recombination but with variable mutation rate along the sequences, and recorded how often the methods take patterns caused by variable mutation rates to be an effect of recombination. Rates of recombination and mutation are varied within realistic values.
