ACRES: efficient query answering on large compressed sequences

Bin Wang, Xiaochun Yang, Guoren Wang

Abstract: With the advances in next generation sequencing, the amount of genomic sequence data being produced continues to grow at an exponential rate. It is estimated that the entire genome of each individual human, each containing about 3 billion letters, could be made available in the next a few years. An increasingly pressing issue in genomics and medicine is how to efficiently store and query these massive amounts of sequence data. Recently a lossless compression technique has been proposed to drastically reduce the storage space of genomic sequences, taking advantage of the fact that any two genomes from the same species are highly similar and therefore only their differences need to be encoded. In this paper we study how to efficiently answer queries on the compressed sequences without first decompressing them. We study three important types of queries, including retrieving a subsequence, finding subsequences matching a given pattern, and finding subsequences similar to a pattern. We propose an index structure, filtering techniques, and efficient algorithms for answering these queries. We further demonstrate the utility of these algorithms using a real dataset.