A parallel approach to multiple sequences alignment and phylogenetic tree node labelling.

摘要

An evolutionary tree represents the relationship among a group of species or sequences. The quality of tree topology relies heavily on an efficient Multiple Sequence Alignment (MSA). Efficient and parallel algorithms are designed to utilise the computing power and memory in a supercomputer. A divide and conquer based parallel algorithm is implemented to perform optimal three sequence alignment with much reduced memory cost. All internal nodes generated from a parallel Maximum Likelihood tree software are labelled using our level order based parallel approach. Such node labelling process is also parallelised to lead to a two-level nested parallel computing strategy.