GPU accelerated suffix array construction for large genome sequences

Next generation sequencing technologies have made it possible to sequence the genome of an organism in hours. Genome is basically a string made of four amino acids represented as A, C, G and T. Only these four amino acids make up for the complete DNA sequence of an organism. These genome sequences are huge in size and requires efficient techniques to analyze them. Well known data structures used to analyze genomes include suffix array/tree, fm-index, burrows wheeler transform etc. In computational biology, pattern matching is often used to efficiently detect the presence of a gene sequence within the genome sequence. This type of sequence analysis requires very fast methods to construct suffix array for the genome sequences. In this paper we have proposed a GPU based method to construct suffix arrays efficiently. We have constructed suffix arrays for homo sapiens chromosomes of GRCh38.p7 primary assembly on varied sizes of upto 1.4GB and we have achieved a significant speed up of upto 3x against the best sequential implementation.