SEGN: Inferring real-time gene networks mediating phenotypic plasticity

Abstract

The capacity of an organism to alter its phenotype in response to environmental perturbations changes over developmental time and is a process determined by multiple genes that are co-expressed in intricate but organized networks. Characterizing the spatiotemporal change of such gene networks can offer insight into the genomic signatures underlying organismic adaptation, but it represents a major methodological challenge. Here, we integrate the holistic view of systems biology and the interactive notion of evolutionary game theory to reconstruct so-called systems evolutionary game networks (SEGN) that can autonomously detect, track, and visualize environment-induced gene networks along the time axis. The SEGN overcomes the limitations of traditional approaches by inferring context-specific networks, encapsulating bidirectional, signed, and weighted gene-gene interactions into fully informative networks, and monitoring the process of how networks topologically alter across environmental and developmental cues. Based on the design principle of SEGN, we perform a transcriptional plasticity study by culturing Euphrates poplar, a tree that can grow in the saline desert, in saline-free and saline-stress conditions. SEGN characterize previously unknown gene co-regulation that modulates the time trajectories of the trees’ response to salt stress. As a marriage of multiple disciplines, SEGN shows its potential to interpret gene interdependence, predict how transcriptional co-regulation responds to various regimes, and provides a hint for exploring the mass, energetic, or signal basis that drives various types of gene interactions.

Characterization, costs, cues and future perspectives of phenotypic plasticity

“…The systems evolutionary game network is a computational model that reconstructs temporal variation in gene networks and track real-time changes in network architecture underlying plastic responses. This approach was used to identify and characterize gene co-regulation that modulates the time trajectories of poplar tree responses to salt stress ( Jiang et al , 2020 ). The molecular mechanisms underlying plastic responses are generally not well understood, but new frameworks and technological advancements will enable new insights into this area.…”

Heterophylly Quantitative Trait Loci Respond to Salt Stress in the Desert Tree Populus euphratica

“…To explore whether heterophylly is related to salt tolerance and whether salinity affects the variation of the heterophylly in P. euphratica, we used these materials to create a control group covering all 106 genotypes, and a salt treatment group covering 86 genotypes, which contained more clonal plants. According to a previous study (Jiang et al, 2020), the control and salt treatment groups were treated with clean water and 300 mM NaCl every 5 days. The NaCl concentration for the salt treatment group was determined according to the methods described in previous studies.…”

Modeling genome-wide by environment interactions through omnigenic interactome networks

“…The additional advantage of functional mapping lies in its capacity to interpolate as many time points as needed for precise network reconstruction. Jiang et al (2020) performed a computer simulation study to investigate the statistical properties of the nLV-based network model, in comparation with dynamic Bayesian networks (Lè bre, 2009) and sparse additive ODEs (Wu et al, 2014). With the same number of time points and the same measurement error, the nLV model performs better in terms of parameter estimation precision, power, and false discovery rate (FDR) than the latter two existing approaches.…”

“…With the same number of time points and the same measurement error, the nLV model performs better in terms of parameter estimation precision, power, and false discovery rate (FDR) than the latter two existing approaches. By using the simulations strategy of Jiang et al (2020), we simulate a set of data that do not contain dependent effects in ODEs of Equation 5. Thus, if the dependent effects (i.e., edges of the network) are detected to be significant, then this means that a type I error is detected.…”