Incorporating functional inter-relationships into protein function prediction algorithms.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Functional classification schemes, like the Gene Ontology (GO), are crucial for protein function prediction.
• Existing methods often overlook the hierarchical relationships within these schemes.
• The GO's Directed Acyclic Graph (DAG) structure presents both opportunities and challenges for prediction algorithms.

Purpose of the Study:

• To develop a method that leverages the inter-relationships within functional classification schemes to enhance protein function prediction.
• To improve the accuracy of supervised protein function prediction algorithms.

Main Methods:

• Incorporated functional class inter-relationships into a k-nearest neighbor classifier using semantic similarity measures.
• Utilized large genomic datasets for modeling and predicting over 100 classes from the GO Biological Process ontology.
• Integrated information from the entire GO hierarchy to improve prediction accuracy.

Main Results:

• The proposed method significantly improves prediction accuracy for numerous functional classes.
• Classes with fewer members benefited most from the incorporation of hierarchical information.
• Demonstrated the discovery of novel functional annotations for yeast proteins, including Sna4, Rtn1, and Lin1.

Conclusions:

• A novel methodology effectively incorporates functional class inter-relationships into protein function prediction algorithms.
• This approach enhances prediction accuracy and facilitates the discovery of new biological insights.
• The developed methodology and associated resources are available for non-commercial use.