Distributed Cache-Reduction Approach to DNA Sequencing Using a Greedy Algorithm for the Shortest Common Superstring

The Shortest Common Superstring (SCS) is an NP-hard problem in computer science with applications ranging from matrix compression to DNA Sequencing. In sequence assembly, the SCS can determine the precise order of DNA fragments. Previous implementations have adopted sequential approaches to the SCS problem. These approaches do not take advantage of massively parallel architectures and fail to reuse computed overlap values between fragments making them inefficient for large data sets. This paper presents a parallel approach to the SCS problem in DNA sequencing. The proposed method utilizes the concept of distributed cache to compute and store overlap values of fragments, and a reduction procedure to merge fragments with high overlapping values recursively until a superstring is obtained. The proposed method greatly reduces memory footprint and scales across multiple nodes in multi-core architectures.