Interpreting protein abundance in Saccharomyces cerevisiae through relational learning

Area of Science:

• Systems Biology
• Proteomics
• Computational Biology

Background:

• Proteomic profiles offer functional insights into an organism's physiological state.
• Understanding protein abundance regulation is crucial in systems biology.
• Saccharomyces cerevisiae is a well-established model organism with extensive biological data.

Purpose of the Study:

• To develop an explainable method for predicting protein abundances.
• To establish predictive relationships between protein levels, function, and phenotype.
• To leverage existing yeast systems biology data for novel insights.

Main Methods:

• Utilized Datalog database representation of biological knowledge.
• Applied relational learning to generate data descriptors.
• Combined descriptors with supervised machine learning for prediction.
• Validated methodology on specific proteins (His4, Ilv2) and phenotypes.

Main Results:

• Successfully predicted protein abundances in an explainable manner.
• Identified predictive relationships linking protein levels to phenotypes like amino acid accumulation and lifespan.
• Demonstrated the method's capability to quantify biological concepts.

Conclusions:

• The developed methodology provides an explainable approach to predicting protein abundances.
• This method can bridge qualitative biological knowledge with quantitative proteomic data.
• The findings contribute to a deeper understanding of yeast systems biology and physiological regulation.