Defining functional distance using manifold embeddings of gene ontology annotations

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Biology

Background:

• Quantifying functional similarity between proteins remains challenging despite established sequence and structure similarity measures.
• Gene Ontology (GO) provides a framework for annotating protein functions, but direct functional distance metrics are lacking.

Purpose of the Study:

• To introduce manifold embedding techniques for computing functional distances between GO annotations.
• To estimate functional distances between protein domains based on their GO annotations.
• To evaluate the accuracy of these functional distances against sequence, structure, and phylogenetic similarities.

Main Methods:

• Application of manifold embedding techniques to Gene Ontology (GO) functional annotations.
• Calculation of functional distances between protein domains.
• Correlation analysis between functional distances and established similarity measures (sequence, structure, phylogeny).

Main Results:

• Functional distances accurately reflect protein domain relationships, mirroring manual classifications of protein structures into folds and superfamilies.
• The new function space reveals proximity between functionally related protein domains.
• Demonstration of how functional distances provide biological context for structure-function relationships.

Conclusions:

• Manifold embedding offers a robust method for quantifying protein functional distances.
• These functional distances provide insights into evolutionary processes like divergent and convergent evolution.
• The methodology is broadly applicable to various biological investigations, including annotation transfer accuracy and sequence-structure-function relationships.