Parallel accelerators for GlimmerHMM bioinformatics algorithm

In the last decades there is an exponential growth in the amount of genomic data that need to be analyzed. A very important problem in biology is the extraction of the biologically functional genomic DNA from the actual genome of the organisms. There have been proposed many computational biology algorithms that solve the gene finding problem which utilize various approaches; GlimmerHMM is considered one of the most efficient such algorithms. This paper presents two different accelerators for the GlimmerHMM algorithm. One of them is implemented on a modern FPGA platform exploiting the parallelism that reconfigurable logic offers and the other one utilizes a GPU (Graphic Processing Unit) taking advantage of a highly multithreaded operational environment. The performance of the implemented systems is compared against the one achieved when the official distribution of the algorithm is executed on a high-end multi-core server; the speedup initiated, for the most compute intensive part, is up to 200× for the FPGA-based system and up to 34× for the GPU-based system.