GPU-accelerated Kendall distance computation for large or sparse data.

Area of Science:

• Computational biology
• Bioinformatics
• Data science

Background:

• Experimental practices generate large, multidimensional datasets.
• Distance matrix calculation is crucial for data preprocessing but computationally intensive.
• Data sparsity presents algorithmic challenges in various fields, including single-cell sequencing.

Purpose of the Study:

• To develop a GPU-enhanced package for fast computation of Kendall-tau distance matrices.
• To address computational demands and data sparsity challenges in large datasets.

Main Methods:

• Developed GPU-Assisted Distance Estimation Software (GADES) for massively parallel Kendall-tau distance computation.
• Implemented specific memory management to overcome GPU memory limitations.
• Incorporated algorithmic solutions to handle data sparsity and enhance acceleration.

Main Results:

• GADES demonstrated significantly higher speed compared to CPU-based packages on simulated and real single-cell sequencing data.
• The software efficiently processes both sparse and dense datasets.
• An additional performance boost was observed for sparse data processing.

Conclusions:

• GADES significantly advances computational strategies for high-performance Kendall distance matrix computation.
• The software enables efficient Big Data processing leveraging GPU power.
• GADES is publicly available for use and further development.