FastTENET: an accelerated TENET algorithm based on manycore computing in Python.

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Gene regulatory networks (GRNs) are crucial for understanding cellular function.
• Single-cell RNA sequencing (scRNAseq) enables high-resolution GRN inference.
• Existing methods like TENET face computational challenges with large scRNAseq datasets.

Purpose of the Study:

• To develop a computationally efficient algorithm for GRN reconstruction from scRNAseq data.
• To accelerate the analysis of large-scale single-cell gene expression datasets.
• To improve the scalability of transfer entropy-based GRN inference.

Main Methods:

• Developed FastTENET, an array-computing implementation of the TENET algorithm.
• Optimized FastTENET for parallel processing on manycore architectures, including GPUs.
• Utilized unique pattern counting of joint events for transfer entropy computation.

Main Results:

• FastTENET demonstrates substantial performance improvements over the original TENET algorithm.
• Achieved up to 973x speedup in GRN reconstruction for large scRNAseq datasets.
• Successfully applied array computing principles for efficient TE calculation.

Conclusions:

• FastTENET overcomes the computational limitations of TENET for scRNAseq data analysis.
• The proposed method enables faster and more scalable GRN inference.
• GPU acceleration significantly enhances the efficiency of transfer entropy-based network reconstruction.