This article is part of the supplement: Selected papers from the Seventh Asia-Pacific Bioinformatics Conference (APBC 2009)
Minimizing recombinations in consensus networks for phylogeographic studies
1 Computational Biology Center, IBM T J Watson Research, Yorktown, USA
2 Department of Computer Science, Rensselaer Polytechnic Institute, New York, USA
3 Biologia Evolutiva, Universitat Pompeu Fabra, Barcelona, Catalonia, Spain
4 Work done during an internship at IBM T J Watson Research Center
BMC Bioinformatics 2009, 10(Suppl 1):S72 doi:10.1186/1471-2105-10-S1-S72Published: 30 January 2009
We address the problem of studying recombinational variations in (human) populations. In this paper, our focus is on one computational aspect of the general task: Given two networks G1 and G2, with both mutation and recombination events, defined on overlapping sets of extant units the objective is to compute a consensus network G3 with minimum number of additional recombinations. We describe a polynomial time algorithm with a guarantee that the number of computed new recombination events is within ϵ = sz(G1, G2) (function sz is a well-behaved function of the sizes and topologies of G1 and G2) of the optimal number of recombinations. To date, this is the best known result for a network consensus problem.
Although the network consensus problem can be applied to a variety of domains, here we focus on structure of human populations. With our preliminary analysis on a segment of the human Chromosome X data we are able to infer ancient recombinations, population-specific recombinations and more, which also support the widely accepted 'Out of Africa' model. These results have been verified independently using traditional manual procedures. To the best of our knowledge, this is the first recombinations-based characterization of human populations.
We show that our mathematical model identifies recombination spots in the individual haplotypes; the aggregate of these spots over a set of haplotypes defines a recombinational landscape that has enough signal to detect continental as well as population divide based on a short segment of Chromosome X. In particular, we are able to infer ancient recombinations, population-specific recombinations and more, which also support the widely accepted 'Out of Africa' model. The agreement with mutation-based analysis can be viewed as an indirect validation of our results and the model. Since the model in principle gives us more information embedded in the networks, in our future work, we plan to investigate more non-traditional questions via these structures computed by our methodology.