Learning Petri net models of non-linear gene interactions

Area of Science:

• Computational Biology
• Genetics
• Systems Biology

Background:

• Understanding genetic influences on disease risk is crucial.
• Machine learning can identify genotype-disease relationships, but biochemical model building is less explored.
• Non-linear gene-gene interactions contribute to complex diseases.

Purpose of the Study:

• To develop an automated method for constructing biochemical models of disease-causing gene-gene interactions.
• To utilize Petri nets for modeling dynamic, concurrent biological processes.
• To validate the method's efficacy in identifying known disease-gene interactions.

Main Methods:

• A random hill climbing algorithm is employed for automated model construction.
• Petri nets are used as the modeling formalism, suitable for biochemical networks.
• The method is tested on three recently reported disease-gene interactions.

Main Results:

• The developed method automatically builds Petri net models.
• It effectively captures non-linear and multi-factorial gene-gene interactions.
• The approach successfully identified perfect Petri net models for all three tested disease-gene interactions.

Conclusions:

• The automated Petri net modeling approach is effective for deciphering complex gene-gene interactions in disease.
• This method provides a valuable tool for understanding the biochemical basis of genetic disease risk.
• The findings highlight the potential of machine learning in building explanatory models of biological systems.