ispace-log

ispace-log infers gene-gene networks (gene regulatory networks) from high-dimensional gene expression data using a log penalty to identify conditional dependencies between genes.

Key Features:

• Log Penalty Integration: Incorporates a log penalty that spans the range between L0 and L1 penalties and improves variable selection consistency compared with L1 (lasso), L2 (ridge), and elastic net penalties.
• SPACE Model Framework: Integrates the log penalty into the Sparse PArtial Correlation Estimation (SPACE) framework to estimate sparse partial correlations for network identification.
• Hub-aware Network Inference: Enhances inference accuracy for networks with complex structures, including networks characterized by hub genes.
• Computational Efficiency: Implemented in C to support efficient computation on large-scale gene expression datasets.

Scientific Applications:

• Gene Regulatory Network Reconstruction: Reconstructs gene regulatory networks (GGNs) by identifying conditional dependencies between genes from expression data.
• Mechanistic Investigation: Facilitates analysis of gene interaction patterns to investigate mechanisms of gene expression regulation and interactions relevant to complex biological processes and diseases.

Methodology:

Performs penalized regression using a log penalty within the SPACE model to obtain sparse partial correlation estimates, implemented in C.