Figure 2.

(A) Four “protein” sequences containing three “domains” (A, B, and C; shown as boxes) and unique regions (shown as lines). (B) A row–column multiple alignment introduces gaps (dotted lines) to align domains A and C but cannot represent the alignment of all three domains. (C) The POA graph improves the alignment in B by reducing the number of gaps but does not align all copies of the domains. (D) A representation of the domain structure as a graph with cycles. (E) We obtain a representation of the multiple alignment of the four sequences by “gluing” together similar regions in the sequences. However, the sequences do not align over their entire length, and the shuffled domains create cycles in the resulting graph. (F) A simplified representation of the ABA graph shows the domains as edges of high multiplicity, and the unaligned regions as edges of multiplicity one.