NIR

NIR infers genetic regulatory networks by solving per-gene multiple linear regression problems with a fixed number of k regressors to identify direct gene–gene interactions from transcriptional perturbation data.

Key Features:

• Linear Regression Framework: Formulates identification of gene interactions as a multiple linear regression problem assuming network sparsity with a fixed number k of regressors per gene.
• Regressor Selection via RSS Minimization: Selects regressors by minimizing the residual sum of squares (RSS) to identify k-tuples that best fit observed expression data.
• Exhaustive Search for Optimal Regressors: Performs exhaustive search across all possible k-tuples of genes to identify regressors that minimize RSS.
• Handling Large Networks with Heuristic Search: Employs a heuristic search method to explore regressor space for large networks when exhaustive search is computationally infeasible.
• Robustness to Noise: Operates effectively in the presence of significant experimental noise, maintaining reliable network identification.
• Predictive Capability: Predicts genes that directly mediate the action of a compound based on transcriptional perturbation data.

Scientific Applications:

• Genetic Network Identification: Infers regulatory architecture by modeling genetic interactions as linear differential equations derived from steady-state gene expression measurements following transcriptional perturbations.
• Experimental Feasibility: Applies to experimental workflows that overexpress genes using episomal plasmids and measure resultant mRNA concentration changes.
• Large-Scale Network Analysis: Enables analysis of extensive genomic networks using heuristic search to accommodate computational scale.

Methodology:

Formulates per-gene multiple linear regression with a fixed number k of regressors; selects k-tuples by minimizing the residual sum of squares (RSS) via exhaustive search across k-tuples or via a heuristic search for large networks; models interactions as linear differential equations from steady-state perturbation measurements.