Time-lagged Ordered Lasso for network inference.

Area of Science:

• Systems Biology
• Computational Biology
• Genomics

Background:

• Gene regulatory networks are crucial for understanding biological functions and disease mechanisms.
• Existing methods for inferring these networks from gene expression data face challenges like low sample size and inaccurate dynamic characterizations.
• Time-course data offers potential for inferring causality but is often limited by short sampling durations.

Purpose of the Study:

• To develop and evaluate novel computational methods for accurate de novo reconstruction of gene regulatory networks.
• To enhance network inference by incorporating prior biological information into a robust regression framework.
• To address limitations in existing temporal methods for gene regulatory network inference.

Main Methods:

• Adaptation of the time-lagged Ordered Lasso, a regression method incorporating temporal monotonicity constraints.
• Development of a semi-supervised approach that integrates prior network information into the Ordered Lasso framework.
• Utilizing R for implementation, with code publicly available.

Main Results:

• The adapted time-lagged Ordered Lasso enables de novo reconstruction of gene regulatory networks.
• The semi-supervised method effectively discovers novel regulatory dependencies within existing pathways.
• Demonstrated accuracy in network reconstruction using simulated data, DREAM challenge datasets, and a HeLa cell cycle dataset.

Conclusions:

• The developed methods produce accurate gene regulatory networks by respecting the dynamics and assumptions of the time-lagged Ordered Lasso regression.
• These approaches offer improved capabilities for inferring gene regulatory networks from time-course expression data.
• The findings contribute to more reliable network inference for biological discovery.