Multiple Partial Order Alignment as a Graph Problem

Multiple Sequence Alignment (MSA) is a fundamental tool of bioinformatics. Row-Column MSA (RC-MSA) methods such as CLUSTALW [12] produce tabular alignments that are now familiar. However, these methods have a number of shortcomings, including difficulty of understanding the result, high computational complexity, questionable assumptions, and other artifacts (such as poor handling of prefixand suffix-alignment). Partial Order Alignment was proposed recently as an alternative approach to MSA. Partial Order MSA (PO-MSA) methods produce a partial order — a labeled directed acyclic graph — that includes the input sequences as subgraphs. The approaches differ in their strengths and weaknesses as well as their assumptions. In this paper, we formalize PO-MSA as a graph problem, show that it corresponds to finding a Minimal Common Supergraph for a set of partial order graphs, and characterize how such a supergraph can be derived. This formalization offers some perspective on MSA generally, and also on particular tradeoffs between RC-MSA and PO-MSA.