A Space Efficient Persistent Implementation of an Index for DNA Sequences

Due to newly developed high-throughput technologies for DNA sequencing, the number of fully sequenced species increases rapidly. String databases holding these sequences are very large. On the eld of molecular biology the handling of large string data which cannot be broken in words is a great challenge. Hereby the most important string operation is the approximate substring match. This type of match is essential for many applications in biology. The suÆx tree has been established as the most predestined in-memory data structure supporting approximate substring matches on DNA sequences. They are belonging to the wider class of suÆx structures. The central issue of the paper is the theoretically evaluation of the suÆx structures with the aim to reveal the most suitable structure for a persistent implementation on the disk. I will show that this is a variant of a suÆx tree. Further, the paper addresses the question in which way this data structure can be stored on disk, and how fast access can be achieved. In this connection I want to introduce a space eÆcient implementation by using a tree coding scheme which optimizes disk saving and therefore preventing unneeded disk access. Here the most important design issue lies in the representation of di erent variants of nodes of the tree. I will present an implementation in C programming language and show that the implementation in a rst evaluation step compares favourably to other implementations.