Defining the extent of gene function using ROC curvature

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Protein interactions are crucial for understanding gene function and phenotypes.
• Machine learning methods extend experimental interaction data across genomes and contexts.
• Evaluating the performance and generalizability of these methods is challenging.

Purpose of the Study:

• To propose a novel method for evaluating the generalizability of gene characterizations.
• To introduce Functional Equivalence Classes (FECs) derived from performance curve shapes.
• To provide a data-driven approach for assessing functional annotation context-specificity.

Main Methods:

• Analyzing the shape of Receiver Operating Characteristic (ROC) curves from gene-centric prediction tasks.
• Identifying straight line patterns in ROC curves to define Functional Equivalence Classes (FECs).
• Assessing the presence and characteristics of FECs across various data types and prediction methods.

Main Results:

• Functional Equivalence Classes (FECs) are widespread and can be visually and statistically assessed.
• FECs enable evaluation of the extent and context-specificity of functional annotations.
• Analysis of B cell markers revealed shared primary and tissue-specific secondary markers.
• Identified functional modules spanning large genomic regions, extending Gene Ontology sets up to 40%.

Conclusions:

• The identification of FECs in performance curves offers a new way to characterize gene function.
• This method enhances the robustness of functional gene set definitions.
• FECs provide valuable insights into the structure and context-dependency of biological functions.