GRENITS

GRENITS infers gene regulatory networks and dynamic interactions from time-series gene expression data using Dynamic Bayesian Networks and Gibbs Variable Selection.

Key Features:

• Dynamic Bayesian Networks: Models temporal dependencies and interactions within time-series gene expression data.
• Gibbs Variable Selection: Performs variable selection to identify relevant predictors and reduce model complexity.
• Gibbs sampling: Uses Gibbs sampling to explore posterior distributions of model parameters.
• Model options: Offers four distinct models tailored to experimental conditions and data characteristics.
• Linear Interaction Model: Captures straightforward linear relationships among variables.
• Linear Models with Experimental Noise — Gaussian Noise Model: Accounts for normally distributed experimental noise suitable for replicated datasets.
• Linear Models with Experimental Noise — Student-t Distributed Noise Model: Accommodates heavier-tailed noise distributions to handle outliers and non-Gaussian noise.
• Non-linear Interaction Model: Captures complex, non-linear relationships within biological systems.
• Bioconductor and R integration: Interoperates with Bioconductor and the R environment for compatibility with other bioinformatics packages.

Scientific Applications:

• Gene regulatory network inference: Infers regulatory interactions from time-series gene expression datasets.
• Dynamic process analysis: Characterizes temporal regulation in gene expression, protein interactions, and cellular signaling pathways.
• Noise-robust network reconstruction: Enables robust inference in the presence of experimental noise using Gaussian and Student-t noise models.
• High-throughput genomics: Applies to high-throughput genomics and molecular biology datasets for network reconstruction.

Methodology:

GRENITS applies Bayesian inference via Dynamic Bayesian Networks combined with Gibbs Variable Selection and Gibbs sampling, supporting linear, linear-with-Gaussian-noise, linear-with-Student-t-noise, and non-linear interaction models.