Optimal, efficient reconstruction of root-unknown phylogenetic networks with constrained and structured recombination

• Published in: Journal of computer and system sciences Vol. 70; no. 3; pp. 381 - 398
• Main Author: Gusfield, Dan
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.05.2005
• Subjects: Ancestral recombination graph; Gene-conversion; Molecular evolution; Perfect phylogeny; Phylogenetic networks; Recombination; Recurrent mutation; SNP; Phylogenetic networks; Molecular evolution; Recombination; Gene-conversion; SNP; Perfect phylogeny; Recurrent mutation; Ancestral recombination graph
• ISSN: 0022-0000; 1090-2724
• DOI: 10.1016/j.jcss.2004.12.009

Phylogenetic networks are models of sequence evolution that go beyond trees, allowing biological operations that are not consistent with tree-like evolution. One of the most important of these biological operations is (single-crossover) recombination between two sequences. An established problem (Math. Biosci. 98 (1990) 185; J. Mol. Evol. 36 (1993) 396; Proceedings of the 2003 Workshop on Algorithms in Bioinformatics, Berlin, Germany, Lecture Notes in Computer Science, Springer, Berlin, 2003; J. Math. Biol. 98 (2003) 160; J. Comput. Biol. 8 (2001) 69; Genetics 163 (2003) 375; Genetics 111 (1985) 147) is to find a phylogenetic network that derives an input set of sequences, minimizing the number of recombinations used. No efficient, general algorithm is known for that problem. An efficient algorithm does exist for the problem when the network is constrained to be a “galled-tree”, and the ancestral sequence for the galled-tree is specified in advance (Proceedings of Second CSB Bioinformatics Conference, Los Alamitos, CA, 2003, IEEE Press, New York; J. Bioinform. Comput. Biol. 2(1) (2004) 173; INFORMS J. Comput. 16 (2004) 459). However, the more biologically realistic case is that no ancestral sequence is known in advance, and the only previous algorithmic solution for that case takes exponential time. In this paper we give an efficient solution to the galled-tree problem when no ancestral sequence is known in advance, and show that the solution produced has very strong global optimality properties. We also indicate how these results generalize to other complex biological phenomena such as gene-conversion, lateral gene transfer, hybrid speciation, and back and recurrent mutation.