Using multi-instance hierarchical clustering learning system to predict yeast gene function.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Time-course gene expression datasets capture dynamic biological processes.
• Predicting gene function is crucial but limited by sparse positive annotations in databases like Gene Ontology (GO).
• Existing methods for establishing learning systems from gene annotations require improvement for enhanced predictive power.

Purpose of the Study:

• To propose a novel Multi-Instance Hierarchical Clustering (MIHC) method for establishing a robust learning system.
• To effectively cluster Gene Ontology (GO) annotations for improved gene function prediction.
• To compare the performance of MIHC against other learning system establishment methods.

Main Methods:

• Development of the Multi-Instance Hierarchical Clustering (MIHC) algorithm.
• Clustering of Gene Ontology (GO) annotations to create a structured learning system.
• Validation using multi-label support vector machine and K-nearest neighbor classifiers on yeast time-course gene expression datasets.

Main Results:

• The proposed MIHC method demonstrated strong performance in gene function prediction.
• MIHC effectively established a learnable and effective learning system from GO annotations.
• Comparison showed MIHC's superiority over alternative methods for learning system establishment.

Conclusions:

• The MIHC method provides a valuable approach for enhancing gene function prediction accuracy.
• This method can serve as a guide for annotators and refine existing gene annotations.
• MIHC contributes to a more effective utilization of time-course gene expression data and GO annotations.