Technical University of Crete, Chania, Greece.
Carnegie Mellon University, Pittsburgh, USA.
Mol Ecol Resour. 2021 Oct;21(7):2580-2587. doi: 10.1111/1755-0998.13438. Epub 2021 Jun 19.
Software tools for linkage disequilibrium (LD) analyses are designed to calculate LD among all genetic variants in a single region. Since compute and memory requirements grow quadratically with the distance between variants, using these tools for long-range LD calculations leads to long execution times and increased allocation of memory resources. Furthermore, widely used tools do not fully utilize the computational resources of modern processors and/or graphics processing cards, limiting future large-scale analyses on thousands of samples. We present quickLD, a stand-alone and open-source software that computes several LD-related statistics, including the commonly used r . quickLD calculates pairwise LD between genetic variants in a single region or in arbitrarily distant regions with negligible memory requirements. Moreover, quickLD achieves up to 95% and 97% of the theoretical peak performance of a CPU and a GPU, respectively, enabling 21.5× faster processing than current state-of-the-art software on a multicore processor and 49.5× faster processing when the aggregate processing power of a multicore CPU and a GPU is harnessed. quickLD can also be used in studies of selection, recombination, genetic drift, inbreeding and gene flow. The software is available at https://github.com/pephco/quickLD.
软件工具用于连锁不平衡 (LD) 分析，旨在计算单个区域中所有遗传变异体之间的 LD。由于计算和内存需求与变异体之间的距离呈二次方增长，因此使用这些工具进行长程 LD 计算会导致执行时间长，并且需要增加内存资源的分配。此外，广泛使用的工具不能充分利用现代处理器和/或图形处理卡的计算资源，限制了未来在数千个样本上进行大规模分析。我们介绍了 quickLD，这是一个独立的开源软件，可计算几个与 LD 相关的统计信息，包括常用的 r。quickLD 计算单个区域或任意远距离的遗传变异体之间的成对 LD，内存需求可忽略不计。此外，quickLD 分别实现了 CPU 和 GPU 的理论峰值性能的 95%和 97%，在多核处理器上的处理速度比当前最先进的软件快 21.5 倍，在多核 CPU 和 GPU 的总处理能力被利用时的处理速度快 49.5 倍。quickLD 还可用于选择、重组、遗传漂变、近亲繁殖和基因流的研究。该软件可在 https://github.com/pephco/quickLD 上获得。
