FastSCODE: an accelerated SCODE algorithm for inferring gene regulatory networks on manycore processors.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables high-resolution gene expression analysis.
• Reconstructing gene regulatory networks (GRNs) is crucial for understanding cellular mechanisms.
• Existing methods like SCODE face computational limitations with large scRNA-seq datasets.

Purpose of the Study:

• To develop a computationally efficient algorithm for GRN reconstruction from scRNA-seq data.
• To optimize SCODE for parallel processing on manycore architectures, such as GPUs.
• To significantly reduce the time required for GRN analysis on large-scale datasets.

Main Methods:

• Developed FastSCODE, a batch computing version of the SCODE algorithm.
• Implemented batch computation on multiple gene expression profiles.
• Utilized manycore computing, specifically GPUs, for parameter optimization of a linear ordinary differential equation (ODE) model.

Main Results:

• FastSCODE achieves substantial performance improvements compared to the original SCODE implementation.
• Demonstrated up to a 6000× speedup on the CeNGEN scRNA-seq dataset using multiple GPUs.
• Reduced processing time from approximately one month to 10 minutes for large datasets.

Conclusions:

• FastSCODE overcomes the computational bottlenecks of SCODE for large scRNA-seq datasets.
• The optimized algorithm enables faster and more scalable GRN reconstruction.
• FastSCODE is publicly available, facilitating its application in systems biology research.