Bayesian model selection for the Drosophila gap gene network.

Area of Science:

• Developmental Biology
• Computational Biology
• Genetics

Background:

• The gap gene system is crucial for insect segmentation and embryo patterning.
• Modeling gap gene dynamics uses reaction-diffusion equations, but evaluating competing models remains a challenge.
• Current model development often occurs in isolation, hindering comparative analysis.

Purpose of the Study:

• To demonstrate a Bayesian framework for formal evaluation of competing gene expression models.
• To apply this framework to models of the Drosophila melanogaster gap gene system.
• To compare different hypotheses within the Papatsenko-Levine formalism.

Main Methods:

• Utilized a Bayesian framework for formal model evaluation.
• Employed the Papatsenko-Levine formalism with fractional occupancy for gene activation and cross-regulation.
• Applied parallel tempering samplers to improve Markov chain convergence and Bayes factor estimation.

Main Results:

• The Bayesian approach revealed negative correlations between binding site number and binding affinity in the segmentation pathway.
• Model selection analysis indicated stronger binding affinity for Bicoid compared to other regulatory interactions.
• The analysis did not support the activation of Kruppel by Bicoid.

Conclusions:

• An efficient solver for the Papatsenko-Levine model representation was developed.
• The utility of Bayes factors for evaluating spatial patterning models was demonstrated.
• Improved Markov chain convergence and robust Bayes factor estimates were achieved using parallel tempering.