Related Experiment Video
Updated: Oct 3, 2025
05:47
Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems
Published on: June 13, 2025
650
An Augmented High-Dimensional Graphical Lasso Method to Incorporate Prior Biological Knowledge for Global Network
Yonghua Zhuang1, Fuyong Xing1, Debashis Ghosh1
1Department of Biostatistics and Informatics, University of Colorado Anschutz Medical Campus, Aurora, CO, United States.
Frontiers in Genetics
|February 14, 2022
Summary
We developed augmented high-dimensional graphical Lasso (AhGlasso) to infer biological networks using protein-protein interaction (PPI) data. AhGlasso is faster and more accurate than existing methods for large-scale omics data analysis.
Area of Science:
- Computational Biology
- Bioinformatics
- Systems Biology
Background:
- Biological networks are often inferred using Gaussian graphical models (GGMs) from gene or protein expression data.
- Conventional GGMs often overlook prior knowledge of protein-protein interactions (PPI), limiting network inference accuracy.
- Existing extensions like weighted graphical Lasso (wGlasso) and network-based gene set analysis (Netgsa) have limitations in computational efficiency or handling PPI network weights.
Purpose of the Study:
- To develop a computationally efficient method for biological network inference that incorporates weighted PPI information.
- To address the limitations of existing methods in handling large-scale omics data and weighted PPI networks.
- To improve the accuracy and speed of global network learning by integrating PPI edge weights.
Main Methods:
- Developed augmented high-dimensional graphical Lasso (AhGlasso), an extension of the Netgsa approach.
- Incorporated edge weights from known PPI networks into the graphical model learning process.
- Evaluated AhGlasso's performance against wGlasso and Netgsa using simulated large-scale data and real proteomic data.
Main Results:
- AhGlasso demonstrates superior computational efficiency compared to wGlasso-based algorithms for large-scale data.
- The runtime of AhGlasso is approximately five times faster than weighted Glasso for graph sizes up to 3,000 nodes.
- AhGlasso achieves comparable or better prediction accuracy of node connections than existing methods.
- Application to chronic obstructive pulmonary disease proteomic data shows improved protein network inference.
Conclusions:
- AhGlasso offers a computationally tractable and accurate approach for biological network inference using omics data and weighted PPI information.
- The method significantly improves network inference by effectively leveraging prior biological knowledge.
- AhGlasso provides a valuable tool for large-scale biological network analysis, particularly in complex diseases.
