Optimizing weighted gene co-expression network analysis with a multi-threaded calculation of the topological overlap...

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Biomolecular networks are often modeled as scale-free hierarchical networks.
• Weighted Gene Co-expression Network Analysis (WGCNA) is a method that treats gene co-expression networks as undirected scale-free hierarchical weighted networks.
• The WGCNA R package identifies biological functions by detecting modules (sub-networks) within these networks, with the calculation of the topological overlap matrix (TOM) from the adjacency matrix being a computationally intensive step.

Purpose of the Study:

• To optimize the computationally expensive calculation of the topological overlap matrix (TOM) in WGCNA.
• To develop a multi-threaded algorithm for faster TOM calculation.
• To enhance the applicability of WGCNA for analyzing large-scale biological datasets.

Main Methods:

• The study modified the single-threaded algorithm for TOM calculation into a multi-threaded version.
• Rcpp was utilized to enable R to call a C++ function.
• OpenMP was employed within the C++ function to execute parallel computations for TOM calculation from the adjacency matrix.

Main Results:

• The multi-threaded algorithm significantly reduces the time required for TOM calculation.
• Performance gains are observed on shared-memory multi-processor systems, with calculation time decreasing as the number of CPU cores increases.
• The optimized algorithm demonstrates the potential for WGCNA application on larger datasets.

Conclusions:

• The developed multi-threaded algorithm accelerates WGCNA by optimizing TOM computation.
• This advancement facilitates the analysis of larger gene co-expression networks.
• The approach is applicable to other research fields requiring sub-network identification in similar network structures.