Reconstructing gene-regulatory networks from time series, knock-out data, and prior knowledge

Area of Science:

• Systems Biology
• Computational Biology
• Genomics

Background:

• Gene regulatory networks control cellular processes.
• Computational methods are used to infer these networks from data.
• This study focuses on time series gene expression and gene knock-out data.

Purpose of the Study:

• To compare network reconstruction methods using synthetic data.
• To investigate data requirements and optimal experiments for network inference.
• To assess the impact of prior knowledge and unobserved cellular processes.

Main Methods:

• Comparison of network reconstruction algorithms.
• Utilizing synthetic data from reference networks.
• Analysis of time series gene expression and gene knock-out data.

Main Results:

• Linear Gaussian dynamic Bayesian networks with F-statistic variable selection are suitable for time series data.
• Discrete dynamic Bayesian networks performed poorly due to information loss.
• Short time series under transcription factor knock-out are optimal experiments.
• Estimates for required gene expression data size based on noise levels were provided.
• Prior knowledge benefits are limited to small data sizes.
• Unobserved processes create dependencies indistinguishable from transcriptional regulation.

Conclusions:

• Gene network reconstruction from expression data remains challenging due to data limitations.
• Optimal experiments, data quality/quantity requirements, and reconstruction methods were identified.
• This study provides insights for reverse engineering gene regulatory networks.