Inferring orthologous gene regulatory networks using interspecies data fusion

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Leveraging existing gene regulatory network (GRN) data from model organisms can inform complex species.
• Cross-species GRN inference is valuable for applications in agriculture and medicine.
• Existing methods often lack the ability to jointly infer or transfer GRN information between species.

Purpose of the Study:

• To develop novel Bayesian methods for joint GRN inference across multiple species.
• To enable the leveraging of GRNs from related species to improve network inference.
• To utilize species-specific time-series gene expression data for enhanced network dynamics modeling.

Main Methods:

• Developed two Bayesian frameworks: direct network propagation and hierarchical propagation via a hypernetwork.
• Incorporated graph kernels to capture network similarity across species.
• Employed Gaussian processes to model time-series gene expression data for species-specific dynamics.

Main Results:

• Joint inference and cross-species leveraging significantly improve GRN inference accuracy compared to standalone methods.
• The non-hierarchical approach is optimal for few species; the hierarchical approach suits many species.
• Successfully predicted a novel cell cycle regulatory role for Gas1 in Schizosaccharomyces pombe using Saccharomyces cerevisiae data.

Conclusions:

• The developed Bayesian methods provide a robust framework for cross-species GRN inference.
• These methods enhance the transfer of biological knowledge between related species, facilitating discovery.
• The study demonstrates the utility of these approaches in predicting gene function and regulatory roles.