ScaleSC: a superfast and scalable single-cell RNA-seq data analysis pipeline powered by GPU.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Large-scale single-cell RNA sequencing (scRNA-seq) data analysis is computationally intensive and slow.
• Existing tools face scalability challenges with increasing dataset sizes.

Purpose of the Study:

• To develop a Graphics Processing Unit (GPU)-accelerated tool for efficient large-scale single-cell RNA sequencing data processing.
• To enhance the scalability and speed of scRNA-seq data analysis.

Main Methods:

• Developed ScaleSC, a GPU-accelerated package leveraging CuPy and CUDA on Scanpy and Rapids-singlecell.
• Implemented GPU-optimized algorithms for core scRNA-seq tasks, marker gene identification, and cluster merging.
• Ensured consistency between GPU and Central Processing Unit (CPU) implementations.

Main Results:

• ScaleSC provides over a 20x speedup for scRNA-seq data processing.
• Successfully processed datasets of 10-20 million cells, significantly surpassing previous capacities.
• Overcame memory bottlenecks on a single A100 GPU, enabling analysis of larger datasets.

Conclusions:

• ScaleSC offers a scalable and efficient solution for processing large-scale scRNA-seq data.
• The tool's Scanpy-like syntax lowers the barrier to adoption for existing users.
• ScaleSC significantly advances the computational capacity for single-cell genomics research.