A robust gene regulatory network inference method base on Kalman filter and linear regression

Area of Science:

• Systems Biology
• Bioinformatics
• Genomics

Background:

• Reconstructing gene regulatory networks (GRNs) from high-throughput genomic data is crucial in systems biology.
• Challenges include high dimensionality (many genes), low sample size, and significant noise in gene expression data.

Purpose of the Study:

• To develop a robust method for inferring GRN topology from noisy gene expression data.
• To improve the accuracy of identifying regulatory relationships between genes.

Main Methods:

• A Kalman filter-based method (KFLR) was developed, incorporating prior knowledge for network learning.
• Mutual information was used to remove low-correlation, noisy regulations in an initial phase.
• A hybrid framework combining Bayesian model averaging and linear regression provided a closed-form solution for posterior probabilities of regulatory edges.

Main Results:

• The proposed KFLR method demonstrated improved inference accuracy compared to existing methods.
• KFLR showed enhanced ability to identify true regulatory relationships, even in the presence of noisy data.
• Evaluation results confirmed the method's efficiency and effectiveness.

Conclusions:

• The KFLR method offers a significant advancement in reconstructing gene regulatory networks from noisy gene expression data.
• This approach enhances the reliability of systems biology research by providing more accurate GRN topologies.