DLIGHT - Lateral Gene Transfer Detection Using Pairwise Evolutionary Distances in a Statistical Framework

Abstract. This paper presents an algorithm to detect lateral gene transfer (LGT) on the basis of pairwise evolutionary distances. The prediction is made from a likelihood ratio derived from hypotheses of LGT versus no LGT, using multivariate normal theory. In contrast to approaches based on explicit phylogenetic LGT detection, it avoids the high computational cost and pitfalls associated with gene tree inference, while maintaining the high level of characterization obtainable from such methods (species involved in LGT, direction, distance to the LGT event in the past). We validate the algorithm empirically using both simulation and real data, and compare its predictions with standard methods and other studies.

• Type: Conference
• Year: 2008
• Where: RECOMB
• Authors: Christophe Dessimoz, Daniel Margadant, Gaston H. Gonnet
• Keywords: Computational Biology, Lateral Gene Transfer, Multivariate Normal Theory, Pairwise Evolutionary Distances, RECOMB 2008