Nonparametric identification of regulatory interactions from spatial and temporal gene expression data

Anil Aswani, Soile V. E. Keränen, James Brown, Charless C. Fowlkes, David W. Knowles, Mark D. Biggin, Peter J. Bickel, Claire J. TomlinBMC Bioinformatics 2010 · DOI · OpenReview

Abstract

Understanding how transcription factors (TFs) coordinate to control gene expression is a central challenge in systems biology. This paper introduces NODE, a novel nonparametric approach for identifying gene regulatory interactions by fitting ordinary differential equation (ODE) models directly to spatial and temporal expression data. By moving away from rigid mathematical assumptions, the method captures complex, concentration-dependent regulatory behaviors in Drosophila embryos, outperforming traditional static models by 59%.

Introduction

Traditionally, researchers have used two main approaches to model Gene Regulatory Networks (GRNs):

  1. Static Correlation Models: These look at whether the presence of a TF correlates with the presence of a target gene, but they ignore the dynamics of the process (how a TF affects the rate of change).
  2. Parametric ODE Models: These use differential equations to model dynamics but require researchers to guess the mathematical "shape" of the interaction (e.g., a specific Hill function). If the biological reality doesn't match the chosen formula, the model fails or over-fits.

The authors address this gap by creating a "quasi-genetic" model that is dynamic like an ODE but flexible like a nonparametric statistical tool, allowing the data to dictate the structure of the interactions.

Methodology

The authors call their method NODE, an amalgamation of the Nonparametric Exterior Derivative Estimator (NEDE) and Ordinary Differential Equations (ODEs).

Key technical features include:

  • Local Linearization: Instead of one global, complex formula, NODE treats the regulatory landscape as a collection of local linear models. It estimates the "slope" of regulation at specific concentration points.
  • NEDE Estimator: This statistical tool protects against over-fitting by learning the underlying manifold (the "shape") of the data, ensuring the model doesn't identify weak noise as strong biological effects.
  • Kernel Weighting: It uses an Epanechnikov kernel to weight data from cells with similar TF concentrations, allowing the model to leverage biological symmetries (like the left-right symmetry of an embryo) automatically.

Datasets and Pattern Analysis

The study utilizes high-resolution data from the Berkeley Drosophila Transcription Network Project (BDTNP). The primary target gene is even-skipped (eve) mRNA, famous for its seven-stripe pattern during Stage 5 of Drosophila melanogaster development. The model was trained using five key transcription factors: Krüppel (KR), Giant (GT), Knirps (KNI), Hunchback (HB), and Bicoid (BCD).

The spatial expression pattern of even-skipped (eve) in the Drosophila embryo, showing the 3D representation (A) and the flattened 2D projection (B) with the seven characteristic stripes.

Key Results and Insights

  • Superior Accuracy: The NODE model showed 59% better agreement with experimental measurements of the eve pattern compared to a state-of-the-art static spatial-correlation model.
  • Pattern Reproduction: As shown in the figure above, the model successfully simulated the narrowing and anterior shifting of eve stripes over time, capturing the dynamic "sharpening" of the developmental pattern.
  • Concentration-Dependent Switching: A major finding is that a single transcription factor can act as both an activator and a repressor for the same gene, depending on its local concentration.
  • Dynamics are Essential: The paper demonstrates that the rate of change in mRNA levels contains far more information about regulation than static snapshots. Static models often misidentify the boundaries of gene expression because they cannot account for the time lag in biological processes.

Limitations and Future Work

While NODE is more accurate, its performance dropped at the very end of Stage 5, likely due to the emergence of unmeasured transcription factors. Additionally, nonparametric methods generally require high data density, which may be a challenge for organisms with sparser datasets than the Drosophila "Virtual Embryo." Future work will focus on applying NODE to larger networks and improving the interpretability of its local linear models into general biological rules.

Summary updated

Abstract

Background The correlation between the expression levels of transcription factors and their target genes can be used to infer interactions within animal regulatory networks, but current methods are limited in their ability to make correct predictions. Results Here we describe a novel approach which uses nonparametric statistics to generate ordinary differential equation (ODE) models from expression data. Compared to other dynamical methods, our approach requires minimal information about the mathematical structure of the ODE; it does not use qualitative descriptions of interactions within the network; and it employs new statistics to protect against over-fitting. It generates spatio-temporal maps of factor activity, highlighting the times and spatial locations at which different regulators might affect target gene expression levels. We identify an ODE model for eve mRNA pattern formation in the Drosophila melanogaster blastoderm and show that this reproduces the experimental patterns well. Compared to a non-dynamic, spatial-correlation model, our ODE gives 59% better agreement to the experimentally measured pattern. Our model suggests that protein factors frequently have the potential to behave as both an activator and inhibitor for the same cis-regulatory module depending on the factors' concentration, and implies different modes of activation and repression. Conclusions Our method provides an objective quantification of the regulatory potential of transcription factors in a network, is suitable for both low- and moderate-dimensional gene expression datasets, and includes improvements over existing dynamic and static models.
BibTeX
@article{aswani2010nonparametric,
  author = {Anil Aswani and Soile V. E. Keränen and James Brown and Charless C. Fowlkes and David W. Knowles and Mark D. Biggin and Peter J. Bickel and Claire J. Tomlin},
  title = {{Nonparametric identification of regulatory interactions from spatial and temporal gene expression data}},
  journal = {BMC Bioinformatics},
  year = {2010},
  doi = {10.1186/1471-2105-11-413},
}
Figure