spacelog

spacelog infers gene-gene networks from high-dimensional gene expression data using a log penalty within Sparse Partial Correlation Estimation (SPACE) to improve variable selection consistency.

Key Features:

• Log Penalty Framework: Applies a log penalty that interpolates between L0 and L1 penalties to enhance variable selection consistency compared with L1, L2, or elastic net.
• Sparse Partial Correlation Estimation (SPACE): Implements SPACE to infer conditional dependencies among genes, with network edges representing putative regulatory relationships.
• Performance with Network Hubs: Demonstrates superior performance for networks characterized by hub genes with many connections.
• Computational Implementation: Implemented in C for computational efficiency on large-scale gene expression datasets.

Scientific Applications:

• Gene Regulatory Network Construction: Constructs gene regulatory networks from gene expression data by inferring conditional dependencies.
• Analysis of Hub Genes: Facilitates analysis of networks with hub genes to study complex biological systems.
• High-Dimensional Genomics Studies: Supports high-dimensional gene expression studies requiring improved variable selection consistency.
• Investigation of Regulatory Mechanisms and Disease: Aids investigation of regulatory mechanisms and gene interactions relevant to biological processes and diseases.

Methodology:

Performs Sparse Partial Correlation Estimation with a log penalty (spanning L0–L1) to select variables and infer conditional dependencies, implemented in C.