BADDADAN: Mechanistic modelling of time-series gene module expression

Area of Science:

• Plant biology
• Systems biology
• Computational biology

Background:

• Plants utilize complex gene regulatory networks (GRNs) to respond to environmental stresses such as heat and drought.
• Accurate modeling of large-scale GRNs (>100 genes) is challenging due to the high parameter estimation burden in traditional ordinary differential equation (ODE) models.

Purpose of the Study:

• To develop a novel computational approach, BADDADAN, for modeling large-scale plant GRNs under stress conditions.
• To improve the understanding of plant resilience mechanisms by predicting gene module dynamics.

Main Methods:

• BADDADAN integrates machine learning with mechanistic modeling.
• It identifies gene modules using time-series gene expression and prior co-expression data.
• An ODE model is constructed to predict gene module dynamics.

Main Results:

• BADDADAN successfully modeled over 1,000 genes in *A. thaliana* under heat and drought stress.
• The approach identified coherent and interpretable gene modules.
• Known mechanistic insights were recovered, and new hypotheses were generated.

Conclusions:

• BADDADAN offers a powerful method for analyzing complex GRNs in plants, enhancing our understanding of stress responses.
• This combined machine learning and mechanistic modeling approach can deepen insights into GRNs in plants and potentially other organisms.