Independence screening for high dimensional nonlinear additive ODE models with applications to dynamic gene...

Area of Science:

• Systems Biology
• Computational Biology
• Genomics

Background:

• Ordinary differential equation (ODE) models are crucial for viral dynamics and epidemic modeling.
• Low-dimensional ODEs struggle with molecular-level disease pathogenesis, like transcriptomic and proteomic data.
• Existing linear ODE models for gene regulatory networks (GRNs) are insufficient due to common nonlinear regulations.

Purpose of the Study:

• To develop a method for reconstructing large-scale nonlinear gene regulatory networks (GRNs) from time-course gene expression data.
• To address the challenge of high dimensionality in modeling complex biological systems.
• To enable a deeper understanding of disease pathogenesis at the molecular level.

Main Methods:

• Utilized high-dimensional nonlinear additive ODEs to model GRNs.
• Developed a 4-step procedure for efficient variable selection in nonlinear ODEs.
• Coupled a 2-stage smoothing-based estimation method with nonlinear independence screening to handle high dimensionality.

Main Results:

• The proposed method demonstrates the sure screening property, effectively selecting relevant variables.
• The approach can handle biological problems with non-polynomial dimensionality.
• Successfully identified the dynamic GRN of Saccharomyces cerevisiae using simulated and real data.

Conclusions:

• The developed method offers an efficient way to reconstruct complex, high-dimensional nonlinear GRNs.
• This advancement is critical for understanding molecular mechanisms underlying disease pathogenesis.
• Provides a robust framework for analyzing dynamic biological networks from gene expression data.