Scalable and Versatile k-mer Indexing for High-Throughput Sequencing Data


Philippe et al. (2011) proposed a data structure called Gk arrays for indexing and querying large collections of high-throughput sequencing data in main-memory. The data structure supports versatile queries for counting, locating, and analysing the coverage profile of k-mers in short-read data. The main drawback of the Gk arrays is its space-consumption, which can easily reach tens of gigabytes of main-memory even for moderate size inputs. We propose a compressed variant of Gk arrays that supports the same set of queries, but in both near-optimal time and space. In practice, the compressed Gk arrays scale up to much larger inputs with highly competitive query times compared to its non-compressed predecessor. The main applications include variant calling, error correction, coverage profiling, and sequence assembly.

Bioinformatics Research and Applications
data structure NGS memory