Jambe

Probabilistic Divergence Measures for Detecting Recombination with Markov Chain Monte Carlo

Version 1.0 alpha, September 2000

Synth4

In a first study, we applied the method to a synthetic DNA sequence alignment of 4 species. The evolution process was modelled with the Kimura 2-Parameter model along an internal branch of length 0.2 and four external branches of equal length 0.1. The transition-tranversion ratio was fixed at 2. The total length of the alignment was N=1000 base pairs, and two recombination events were simulated after traversing 80% of the external branches: Strains 2 and 3 exchanged a consecutive patch of 100 basepairs between sites 201 and 300 in the alignment, and strains 2 and 4 exchanged a chunk of 300 basepairs between sites 600 and 900. After these recombination events, the evolutionary process was simulated according to the new tree for the remaining 20% of the external branch lengths. We slided a window of W=100 bp along the alignment in step sizes of Delta W = 10 bp. The Markov chain Monte Carlo simulations were carried out with BAMBE, using the default parameter settings from the parameter file provided with the software. For each position of the window, 15000 Metroplis-Hastings steps were carried out, of which the first 5000 steps were discarded (burn in). Tree configurations were sampled every ten steps, giving a total of 1000 trees for each window. The results thus obtained are shown in the following figure:

The flanking regions of the recombinant zones are marked by peaks in the entropy H(t), which reflects the increased uncertainty about the topology. The recombinant regions are clearly marked in the plots of the two Kullback-Leibler divergence measures.

Back to my homepage
Last modified: 22 September 2000