Toward optimal disk layout of genome scale suffix trees

Abstract

Suffix trees provide for efficient indexing of numerous sequence processing problems in biological databases. We address the pivotal issue of improving the search efficiency of disk-resident suffix trees by improving the storage layout from a statistical learning viewpoint. In particular, we make the following contributions: we (a) introduce the Q-Optimal Disk Layout(Q-OptDL) problem in the context of suffix trees and prove it to be NP-Hard, and (b) propose an algorithm for improving the layout of suffix trees that is guaranteed to perform asymptotically no worse than twice the optimal disk layout. © 2010 Springer-Verlag.