Space-efficient construction of the Burrows-Wheeler transform

Abstract

The Burrows-Wheeler transform (BWT), originally invented for data compression, is nowadays also the core of many self-indexes, which can be used to solve many problems in bioinformatics. However, the memory requirement during the construction of the BWT is often the bottleneck in applications in the bioinformatics domain. In this paper, we present a linear-time semi-external algorithm whose memory requirement is only about one byte per input symbol. Our experiments show that this algorithm provides a new time-memory trade-off between external and in-memory construction algorithms. © Springer International Publishing 2013.