TY - GEN
T1 - A faster algorithm for RNA co-folding
AU - Ziv-Ukelson, Michal
AU - Gat-Viks, Irit
AU - Wexler, Ydo
AU - Shamir, Ron
PY - 2008/11/28
Y1 - 2008/11/28
N2 - The current pairwise RNA (secondary) structural alignment algorithms are based on Sankoff's dynamic programming algorithm from 1985. Sankoff's algorithm requires O(N 6) time and O(N 4) space, where N denotes the length of the compared sequences, and thus its applicability is very limited. The current literature offers many heuristics for speeding up Sankoff's alignment process, some making restrictive assumptions on the length or the shape of the RNA substructures. We show how to speed up Sankoff's algorithm in practice via non-heuristic methods, without compromising optimality. Our analysis shows that the expected time complexity of the new algorithm is O(N 4 ζ(N)), where ζ(N) converges to O(N), assuming a standard polymer folding model which was supported by experimental analysis. Hence our algorithm speeds up Sankoff's algorithm by a linear factor on average. In simulations, our algorithm speeds up computation by a factor of 3-12 for sequences of length 25-250. Availability: Code and data sets are available, upon request.
AB - The current pairwise RNA (secondary) structural alignment algorithms are based on Sankoff's dynamic programming algorithm from 1985. Sankoff's algorithm requires O(N 6) time and O(N 4) space, where N denotes the length of the compared sequences, and thus its applicability is very limited. The current literature offers many heuristics for speeding up Sankoff's alignment process, some making restrictive assumptions on the length or the shape of the RNA substructures. We show how to speed up Sankoff's algorithm in practice via non-heuristic methods, without compromising optimality. Our analysis shows that the expected time complexity of the new algorithm is O(N 4 ζ(N)), where ζ(N) converges to O(N), assuming a standard polymer folding model which was supported by experimental analysis. Hence our algorithm speeds up Sankoff's algorithm by a linear factor on average. In simulations, our algorithm speeds up computation by a factor of 3-12 for sequences of length 25-250. Availability: Code and data sets are available, upon request.
UR - http://www.scopus.com/inward/record.url?scp=56649106364&partnerID=8YFLogxK
U2 - 10.1007/978-3-540-87361-7_15
DO - 10.1007/978-3-540-87361-7_15
M3 - Conference contribution
AN - SCOPUS:56649106364
SN - 3540873600
SN - 9783540873600
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 174
EP - 185
BT - Algorithms in Bioinformatics - 8th International Workshop, WABI 2008, Proceedings
T2 - 8th International Workshop on Algorithms in Bioinformatics, WABI 2008
Y2 - 15 September 2008 through 19 September 2008
ER -