Compressed Suffix Trees for Repetitive Texts

Abstract

Abstract. We design a new compressed suffix tree specifically tailored to highly repetitive text collections. This is particularly useful for sequence analysis on large collections of genomes of the close species. We build on an existing compressed suffix tree that applies statistical compression, and modify it so that it works on the grammar-compressed version of the longest common prefix array, whose differential version inherits much of the repetitiveness of the text.

Faster Compressed Suffix Trees for Repetitive Text Collections

“…However, GCT uses much less space on repetitive collections, around 2-3 bpc. This is slightly larger than the previous structure for repetitive collections [2,1] but one to two orders of magnitude faster than it. On an extremely repetitive collection, GCT uses even less space than that previous structure, near 0.5 bpc.…”

“…Fischer et al [12] proved that H was in addition compressible when the text was statistically compressible, but Cánovas and Navarro [7,1] found out that the compressibility was not significant on standard texts. Instead, Abeliuk and Navarro [2,1] showed that the technique proposed to compress H [12] made a significant difference on repetitive texts.…”

“…Abeliuk and Navarro [2,1] proposed the first CST for repetitive text collections. They build on the representation of Fischer et al, using the RLCSA and the compressed version of H to represent LCP .…”

“…Note that this theoretical idea is what was implemented in practice by Abeliuk and Navarro [2,1], as described in Section 2.2, for solving queries on the LCP array: using the RePair grammar tree instead of a balanced tree for storing m[] and e[] information. In this paper we implement the idea on the excess array of an actual tree -the suffix tree of the text.…”

“…Based on the RLCSA, Abeliuk and Navarro [2,1] introduced the first CST for repetitive collections. The space on the repetitive biological collections tested is around 1-2 bits per character (bpc), well below the spaces achieved with the CSTs for general text collections.…”