Figure 1

Correspondence between alignment columns and edges connecting cells in a dynamic programming matrix, illustrated for pairwise alignment. In order to permit a directed acyclic graph representation of the space of possible alignments, each column is given a code that distinguishes between gaps based upon where they occur in the alignment. The coding for each column for the two alignments shown in panel a) represents a bijection to the midpoints of edges connecting cells in the dynamic programming table in panel b). Cell boundaries are indicated by thicker gridlines, and the finer gridlines indicate the column coding corresponding to each position, as labelled on the top and right axes. These codings are derived from the characters shown on the bottom and left axes. The midpoint of each cell is labelled with a circle, and each edge is annotated with a rectangle denoting the corresponding column. Each path from X ⁽⁰⁾ to X ^(T) (shown as dashed columns at (0,0) and (2,2), respectively) represents a valid alignment.