Fast algorithms for large-scale genome alignment and comparison

Abstract

We describe a suffix-tree algorithm that can align the entire genome sequences of eukaryotic and prokaryotic organisms with minimal use of computer time and memory. The new system, MUMmer 2, runs three times faster while using one-third as much memory as the original MUMmer system. It has been used successfully to align the entire human and mouse genomes to each other, and to align numerous smaller eukaryotic and prokaryotic genomes. A new module permits the alignment of multiple DNA sequence fragments, which has proven valuable in the comparison of incomplete genome sequences. We also describe a method to align more distantly related genomes by detecting protein sequence homology. This extension to MUMmer aligns two genomes after translating the sequence in all six reading frames, extracts all matching protein sequences and then clusters together matches. This method has been applied to both incomplete and complete genome sequences in order to detect regions of conserved synteny, in which multiple proteins from one organism are found in the same order and orientation in another. The system code is being made freely available by the authors.

Figure 1

A sample suffix tree showing the streaming behavior for finding matches between a query and a reference.

Figure 2

Alignment of multiple fragments from the partially sequenced P.yoelii genome to chromosome 2 of P.falciparum. Coordinates on chromosome 2 are displayed along the horizontal axis, and the percent similarity at the amino acid level is show on the vertical axis. Line segments represent regions detected as similar, not entire contigs; thus, a contig may contain multiple, distinct regions of similarity. Percent similarity refers only to the coding regions, not to entire length of each line segment. Contigs from P.yoelii are sorted according to their mapped position on P.falciparum.