A linear programming based algorithm for multiple sequence alignments

In this brief paper we discuss an approach to multiple sequence alignment based on treating the alignment process as a stochastic control problem. The use of a model based on a Markov decision process leads to a linear programming problem whose solution is linked to a suggested alignment. Our goal is to avoid the expense in time and computation encountered when dynamic programming based algorithms are used to align a large number of sequences. The dual linear programming problem can also be defined. We implemented the method on a set of cytochrome p450 sequences and compared our suggested alignments of 3 sequences with that obtained by CLUSTALW.