A look at accelerated hardware in computational biology

Abstract

Welcome to 2014! h FOLLOWING DECEMBER 2013’S special year end issue on the slowing effects of variability and aging in ICs and systems, to start this new volume of D&T we bring you to a completely different space. In this issue, we take a focused look into the growing computational challenges associated with molecular biology research. The generation of biological data is now happening at unprecedented rates, and processing rates have not really kept pace. Such processing has typically been carried out in software on standard desktop computing platforms, but this situation is changing. This issue explores such changes and highlights some of the hardware-based approaches and corresponding algorithms that have been developed to enable highly desired biological data processing acceleration. Our first article, by Majumder et al., dives into the specifics of high-speed rates of data generation in molecular biology research. The article compares the performance of emerging hardware platforms with other applications across the field of computational biology. Second, an article by Aluru and Jammula provides a wider scope investigation by presenting surveys on FPGA and GPU hardware accelerators in biological sequence analysis, as well as research on acceleration resulting from examining high-throughput sequencing and applications. Our third submission, authored by Liu and Schmidt, presents two critical computing techniques for CUDA-enabled GPUs that allow fast alignments to accelerate the CUSHAW2 algorithm, supported by the alignments of simulated and real reads to the human genome. Next, Schlachter et al. investigate the problem of resource utilization in molecular dynamics simulations. They focus on non-dedicated high-end clusters and propose additional modules to supplement existing workflow and resource managers. They report on two molecular simulations and validate the performance benefits that their proposed approaches bring. Following this, authors Savran, Gao, and Bakos discuss improvements to the memory usage and performance of their GPU kernel, which performs large-scale short sequence dataset pair wise alignments. The authors have established a possible new record in large-scale alignments. To conclude this discussion of computational biology acceleration, an article by Chrysos et al. presents a number of informative case studies that exemplify how modern hybrid systems with FPGAbased reconfigurable computing platforms can offer large speed-ups and savings in bioinformatics algorithms. Our last three feature articles touch on more general interest topics. The first is an article by Yilmaz, Nassery, and Ozev that outlines and confirms the accuracy of built-in EVM measurement techniques that use QAM modulation and avoid high DFT overheads. Following this, Jenihhin et. al present an approach to design error localization that combines statistical analysis of VHDL code items with static slicing. The authors demonstrate the efficiency of their approach on an industrial processor (ROBSY) using a set of real bug cases and the original functional test. Our last article, a collaboration by authors from the University of California at Santa Barbara, Aarhus