Extracting Maximal Exact Matches on GPU

Abstract

The revolution in high-throughput sequencing technologies accelerated the discovery and extraction of various genomic sequences. However, the massive size of the generated datasets raise several computational problems. For example, aligning the sequences or finding the similar regions in them, which is one of the crucial steps in many bioinformatics pipelines, is a time consuming task. Maximal exact matches have been considered important to detect and evaluate the similarity. Most of the existing tools that are designed and developed to find the maximal matches are based on advanced index structures such as suffix tree or array. Although these structures triggered the development of efficient search algorithms, they need large indexing tables which yield large memory footprint for the software using them and bring significant overhead. In this article, we introduce a novel tool GPUMEM which effectively utilizes the massively parallel GPU threads while finding maximal exact matches inside two genome sequences using a lightweight indexing structure. The index construction, which is also handled in GPU, is so fast that even by including the index generation time, GPUMEM can be faster in practice than a state-of-the-art tool that uses a pre-built index.