Scholarlink

Abstract <p>The paper discusses an approach to utilizing GPUs for accelerating computations in bioinformatics applications. The study is based on the well-known MATT algorithm used in structural bioinformatics for performing multiple structural alignments of proteins. A method is proposed to offload computations to the GPU for one of the phases of this algorithm—specifically, the phase of selecting pairs of structural fragments for subsequent comparison. This method has previously demonstrated its efficiency on this task when applied to the comparison phase. It is based on the use of computation pipelining when transferring calculations to the GPU. An experimental study of the implemented approach was conducted using the superfamily of proteins from the CATH database. The practical implementation was carried out on the ‘‘Lomonosov-2’’ supercomputer node at Moscow State University. The results demonstrated an acceleration of computations both for the considered computation phase and for the MATT algorithm using the modified phase.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Approach to Accelerate Bioinformatics Algorithms for Proteins 3D Structure Alignment Using Graphics Processing Units

  • 期刊论文

  • 发表日期：

    2025年8月1日

查看全文
  • I. A. Timokhin,
  • N. N. Popova

摘要

Abstract

The paper discusses an approach to utilizing GPUs for accelerating computations in bioinformatics applications. The study is based on the well-known MATT algorithm used in structural bioinformatics for performing multiple structural alignments of proteins. A method is proposed to offload computations to the GPU for one of the phases of this algorithm—specifically, the phase of selecting pairs of structural fragments for subsequent comparison. This method has previously demonstrated its efficiency on this task when applied to the comparison phase. It is based on the use of computation pipelining when transferring calculations to the GPU. An experimental study of the implemented approach was conducted using the superfamily of proteins from the CATH database. The practical implementation was carried out on the ‘‘Lomonosov-2’’ supercomputer node at Moscow State University. The results demonstrated an acceleration of computations both for the considered computation phase and for the MATT algorithm using the modified phase.