Performance extraction and suitability analysis of multi- and many-core architectures for next generation sequencing secondary analysis

Abstract

High-throughput next generation sequencers (NGS) can rapidly read billions of short DNA fragments, called reads, at low cost. Moreover, their throughput is increasing and cost is decreasing at rates much faster than the Moore's law. This demands commensurate acceleration for NGS secondary analysis that process the reads to identify variations between genomes. Conventional architectural improvements can at best improve performance at the rate of Moore's law even if the software tools efficiently utilize the underlying architecture. Unfortunately, most of the dozens of software products developed for this purpose fail to exploit the underlying architecture well. Therefore, to match the pace of development of the sequencers, we will need architecture that is more tailored for the computational requirements of NGS secondary analysis as well as software that uses the architecture optimally. To this end, in this work, we study the performance characteristics of NGS secondary analysis and investigate the suitability of modern Intel Xeon and Xeon Phi processors for the same. To keep the study manageable, we rely on recent studies that attribute a majority of the run-time to a few key kernels. We present detailed optimization efforts to accelerate these kernels on the latest Intel Xeon and Xeon Phi processors with the goal of extracting maximum performance. A comparison of our optimized implementations, along with published results on GPGPU implementations, shows that our * Kanak Mahadik was a research intern at Intel when she worked on this project.

LISA: A Case For Learned Index based Acceleration of Biological Sequence Analysis

“…We demonstrate the efficacy of LISA by comparing the throughput (million-reads/sec) with FM-Index based exact search and SMEM search. For the baseline comparison, we use Trans-Omics Acceleration Library (TAL) which provides the architecture optimized implementations for traditional FM-index exact search and SMEM search [18,21,22]. The optimized SMEM kernel from TAL is also used in BWA-MEM2 [21], an [1] architecture-optimized implementation of BWA-MEM [4].…”

“…The key idea behind an FM-index is that, in the lexicographically sorted order of all suffixes of the reference sequence, all matches of a short DNA sequence (a.k.a., a "query") will fall in a single region matching the prefixes of contiguously located suffixes. Over the years, many improvements have been made to make the FM-index more efficient, leading to several state-of-the-art implementations that are highly cache-and processor-optimized [5,[7][8][9][10][12][13][14][15][16][17][18][19]. Hence, it becomes increasingly more challenging to further improve this critical step in the genomics pipeline to scale with increasing data growth.…”

LISA: Towards Learned DNA Sequence Search

“…Implementation of these updates may make these results inapplicable to your device or system. significantly faster than its alternatives [15]. We also compare with doing backwards search using IP-BWT and binary search (i.e., without the RMI).…”

“…When evaluated on an Intel R Core TM i9-9900K 3.6 GHz processor, despite being single-threaded and not yet fully optimized to the underlying hardware architecture, our current implementation achieves nearly 4× speedup against a state-of-the-art single-threaded, CPU-optimized version of the FM-index based algorithm [15], for a workload of 50 million queries matched against the human genome. This early result shows that learned DNA sequence search is a promising idea 1 .…”

Accelerating Sequence Alignment to Graphs

“…Computing score matrices is highly computeintensive, and consumes most of the time in sequential as well as our parallel algorithm. The proposed algorithm is inspired from previous optimization efforts targeted towards accelerating Smith-Waterman alignment using SIMD instructions [31], [32]. Alignment of a single sequence is called a task.…”