SRGS

SRGS infers gene regulatory networks (GRNs) from bulk and single-cell expression data using a Sparse Partial Least Squares (SPLS)-based recursive gene selection approach to identify candidate regulators and mitigate single-cell RNA-sequencing dropouts.

Key Features:

• Sparse Partial Least Squares (SPLS) Approach: Employs Sparse Partial Least Squares (SPLS) to select and score genes that potentially regulate target genes, accommodating high-dimensional expression data.
• Recursive Gene Selection: Iteratively selects genes based on their regulatory potential with respect to specific target genes to refine inferred GRNs.
• Robustness to Data Dropouts: Enhances robustness to single-cell RNA-seq dropouts by randomly scrambling samples, setting some expression matrix values to zero, and generating multiple perturbed dataset copies through iterative processes.
• Versatility Across Data Types: Applied to simulated bulk data, simulated single-cell data with and without dropouts, and experimental single-cell datasets.
• Benchmarking and Validation: Benchmarked against existing GRN inference methods on diverse simulated and experimental expression datasets.

Scientific Applications:

• Cellular regulatory mechanism discovery: Infer regulatory interactions to elucidate molecular mechanisms underlying cellular processes at bulk and single-cell resolution.
• Disease mechanism investigation: Identify gene regulatory relationships relevant to the molecular basis of complex diseases.
• Single-cell regulatory heterogeneity analysis: Characterize cell-to-cell regulatory variation to inform studies of individual cell behavior and personalized medicine.

Methodology:

Handles bulk and single-cell expression data; addresses single-cell dropout by randomly scrambling samples, zeroing values in the expression matrix, and generating multiple dataset copies; uses SPLS to recursively select and score genes for target-specific regulatory influence; and benchmarks inferred networks against existing GRN inference methods.