Fast algorithms for aligning sequences with restricted affine gap penalties

Abstract

Affine gap penalties are generally considered appropriate for aligning DNA and protein sequences. (“Affine” means that a gap of length k is penalized α + k β, i.e., it costs α to open up a gap plus β for each symbol in the gap.) For certain applications, such as aligning a cDNA sequence with a genomic DNA sequence, it might be adequate to use the restricted affine gap penalties which penalize long gaps with a constant penalty. As it turns out, several techniques developed for solving the approximate string matching problem can be utilized to yield efficient algorithms for computing the optimal alignment with restricted affine gap penalties. In particular, efficient algorithms can be derived based on the suffix automaton with failure transitions and on the diagonalwise monotonicity of the cost tables. To speedup the computation, the q-gram paradigm can be used to locate the interval in the longer sequence that should be aligned with the shorter sequence. We have implemented the above methods in C on Sun workstations running SunOS Unix. Preliminary experiments show that these approaches are very promising for aligning a cDNA sequence with a genomic DNA sequence.

This work was supported in part by grant R01 LM05110 from the National Library of Medicine, National Institutes of Health, USA, and grant NSC86-2213-E-126-002 from the National Science Council, Taiwan.