Global protein function annotation through mining genome-scale data in yeast Saccharomyces cerevisiae

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• The post-genomic era faces challenges in extracting biological knowledge from high-throughput data.
• Accurate protein functional annotation is crucial for understanding biological systems.

Purpose of the Study:

• To develop a robust method for protein functional annotation in yeast (Saccharomyces cerevisiae).
• To integrate diverse high-throughput biological data for improved annotation accuracy.

Main Methods:

• Developed a Bayesian statistical method combined with Boltzmann machine and simulated annealing.
• Constructed a 'functional linkage graph' quantifying protein functional similarity based on integrated data.
• Incorporated evolutionary and subcellular localization information into the prediction model.

Main Results:

• Successfully assigned functions to 1802 out of 2280 unannotated yeast proteins.
• Demonstrated the effectiveness of integrating multiple high-throughput datasets (yeast two-hybrid, protein complexes, gene expression).
• The functional linkage graph effectively represents Bayesian probabilities of function similarity.

Conclusions:

• The developed Bayesian approach offers a systematic and effective solution for protein functional annotation.
• Integration of multiple data types significantly enhances the accuracy of functional prediction.
• This method advances the systematic characterization of biological systems in the post-genomic era.