A generalized framework for controlling FDR in gene regulatory network inference.

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Gene regulatory network (GRN) inference from perturbation data offers mechanistic insights into biological systems.
• Existing GRN inference methods are often sensitive to parameter choices, leading to potential inaccuracies (missing or incorrect links).
• A robust method is needed to estimate the reliability of inferred GRN links across varying parameters.

Purpose of the Study:

• To develop a novel method for estimating the support of predicted links in GRNs.
• To improve the reliability and accuracy of GRN inference.
• To provide a general approach for controlling the false discovery rate in GRN inference.

Main Methods:

• Developed 'nested bootstrapping,' a protocol applying bootstrapping to GRN inference to assess link stability.
• Utilized shuffled data to translate bootstrap support values into false discovery rates.
• Integrated the method into the GeneSPIDER package for simulation, inference, and analysis.

Main Results:

• Nested bootstrapping demonstrated improved inference accuracy across various parameters, noise levels, and data properties.
• The method effectively controls the false discovery rate for GRN inference.
• Evaluated on simulated data using multiple inference algorithms (LASSO, Least Squares, RNI, GENIE3, CLR).

Conclusions:

• Nested bootstrapping offers a general and effective strategy for enhancing the reliability of gene regulatory network inference.
• This approach provides a reliable way to assess the confidence of inferred gene-gene interactions.
• The method is broadly applicable to diverse GRN inference scenarios and algorithms.