Stochastic dynamic modeling of short gene expression time-series data.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Gene regulatory networks (GRNs) are crucial for understanding cellular processes.
• Modeling GRNs from gene time-series data presents challenges due to noise and data sparsity.
• Accurate GRN inference is essential for biological discovery.

Purpose of the Study:

• To apply the expectation maximization (EM) algorithm for modeling gene regulatory networks.
• To simultaneously identify model parameters and gene expression levels from time-series data.
• To demonstrate the EM algorithm's efficacy with noisy and sparse microarray data.

Main Methods:

• Utilizing the expectation maximization (EM) algorithm for stochastic dynamic modeling.
• Modeling gene regulation as a first-order autoregressive (AR) stochastic dynamic process.
• Handling gene expression data with a high number of variables and a small number of observations.

Main Results:

• The EM algorithm successfully models gene regulatory networks using time-series data.
• Simultaneous identification of model parameters and gene expression levels was achieved.
• The algorithm efficiently handles sparse parameter identification and noisy gene expression data.
• Demonstrated effectiveness on four real-world gene expression datasets.

Conclusions:

• The EM algorithm is a robust method for inferring gene regulatory networks.
• The proposed method is suitable for analyzing complex microarray gene expression datasets.
• The developed models provide insights into biological properties of inferred gene regulatory networks.