Restoring information in aged gene regulatory networks by single knock-ins

TL;DR

The paper addresses how aging erodes information in gene regulatory networks and asks whether single-gene perturbations can restore information flow. It introduces a minimal binary information-transmission framework for TF–TG interactions, estimates model parameters from aging data without fitting, and uses knock-in perturbations to predict changes in mutual information $I$ across the network. The main findings show that single knock-ins can restore up to about 10% of the lost information, with greater restoration when effects propagate through the network, and identify top restorative genes such as Ppara, Phox2b, Esrra, Med23, and Ppargc1b. This framework provides a predictive tool for identifying rejuvenation targets and suggests that combinatorial perturbations could yield additive improvements in information flow across gene networks in aging tissues.

Abstract

A hallmark of aging is loss of information in gene regulatory networks. These networks are tightly connected, raising the question of whether information could be restored by perturbing single genes. We develop a simple theoretical framework for information transmission in gene regulatory networks that describes the information gained or lost when a gene is "knocked in" (exogenously expressed). Applying the framework to gene expression data from muscle cells in young and old mice, we find that single knock-ins can restore network information by up to 10%. Our work advances the study of information flow in networks and identifies potential gene targets for rejuvenation.

Figures (6)

• Figure 1: Schematic of a gene regulatory network (blue) in which one gene is perturbed (yellow) by being "knocked in," or exogenously expressed. The knock-in affects the gene's regulatory targets, and their targets, and so on (orange), suggesting that one perturbation could significantly affect information transmission across the whole network.
• Figure 2: Gene expression and information transmission decrease with age. (a) Histograms of processed transcript numbers from the TMS database tabula2020single for limb muscle cells of young (3 months) and old (24 months) mice. Transcript numbers are averaged over 1,102 (young) or 1,232 cells (old) and binned across 2,362 genes. Curves: Gaussian kernel density estimates to guide the eye. Inset: Average and standard error of each histogram. (b) Mutual information (Eq. \ref{['mi']}), averaged over all 6,253 TF-TG pairs, for young and old mice. Error bars: standard error.
• Figure 3: Model of binary gene regulation. (a) A transcription factor (TF) regulates a target gene (TG). (b) Rates determine transitions between the four possible states.
• Figure 4: Information vs. on-probability for two example knocked-in genes. Each curve is the average and standard error of $I$ (Eq. \ref{['mi']}) vs. $p_{\rm on}$ (Eqs. \ref{['ponTF']}, \ref{['ponTG']}) for all regulatory pairs a given distance from the gene in the regulatory network. The blue curves are discussed in Sec. \ref{['one']}, and all curves are discussed in Sec. \ref{['prop']}.
• Figure 5: Maximal fraction of information restored by knock-in, and the on-probability at which it occurs, for the 100 most restorative genes. (a) Considering only regulations in which the knocked-in gene participates (distance 1). (b) Propagating the knock-in to the whole network (all distances). In b, 10 repeats are performed (see Sec. \ref{['prop']}); standard error of resulting information values is smaller than the data points.