Exploring Approaches for Detecting Protein Functional Similarity within an Orthology-based Framework

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein functional similarity (FS) is crucial for applications like gene prioritization and disease gene discovery.
• Existing FS methods rely on Gene Ontology (GO) annotations and semantic similarity (SS) measures.
• Current FS values are difficult to compare directly due to annotation bias and variability.

Purpose of the Study:

• To introduce a novel similarity z-score for more reliable protein functional similarity assessment.
• To evaluate the effectiveness of z-scores in improving FS calculations across different SS measures and mixing strategies.
• To provide a practical tool for high-throughput protein FS calculation and interpretation.

Main Methods:

• Calculating GO term semantic similarity (SS) between protein annotations.
• Applying a 'mixing strategy' to combine SS values into a final FS score.
• Introducing and applying a similarity z-score that accounts for individual protein FS background distributions.
• Benchmarking the z-score approach against standard methods using orthologous and random gene pairs.

Main Results:

• The similarity z-score demonstrated moderate accuracy improvements in distinguishing orthologous gene pairs from random pairs.
• The impact of different annotation corpus choices on z-score performance was analyzed.
• The Frela web server was developed for fast, high-throughput protein FS calculation.

Conclusions:

• The z-score approach offers a more robust and reliable method for quantifying protein functional similarity.
• This method helps mitigate issues arising from protein annotation variability and bias.
• Frela provides a valuable resource for researchers needing to calculate and interpret protein functional similarity.