CAJ | 학술논문

针对分子对接中生成评分网格需要花费很多的计算时间这一问题, 提出了一种基于统一计算设备架构（CUDA）的评分网格生成并行算法。该算法把传统计算方法中三维计算空间中的一维通过在图形处理单元（GPU）上进行并行处理, 使得总生成时间得到了降低, 提高了评分网格的生成效率。实验结果表明, 借助于GPU的浮点计算能力, 提出的并行算法对比传统的计算方法可以显著缩短评分网格的生成时间, 为评分网格的生成提供一种新的方式。
침대분자대접중생성평분망격수요화비흔다적계산시간저일문제, 제출료일충기우통일계산설비가구（CUDA）적평분망격생성병행산법。해산법파전통계산방법중삼유계산공간중적일유통과재도형처리단원（GPU）상진행병행처리, 사득총생성시간득도료강저, 제고료평분망격적생성효솔。실험결과표명, 차조우GPU적부점계산능력, 제출적병행산법대비전통적계산방법가이현저축단평분망격적생성시간, 위평분망격적생성제공일충신적방식。
The time of generating the scoring grid in molecule docking is very huge. Aiming at this problem, this paper introduced a parallel algorithm of score grid based on the compute unified device architecture (CUDA). In traditional methods, it needed to calculate every node in three-dimensional space. This algorithm took the calculation of one dimension on graphic processing unit (GPU). It reduced the total computational time. The experimental results show that the parallel algorithm with the help of GPU, can significantly shorten generation time of the score grid compared with the traditional calculating method of DOCK. The algorithm offers a new way for generating the scoring grid.