Quantifying structure-function uncertainty: a graph theoretical exploration into the origins and limitations of...

Area of Science:

• Structural genomics
• Bioinformatics
• Computational biology

Background:

• Protein domain annotation databases (CATH, SCOP, FSSP) are crucial for structural genomics.
• Accurate classification and annotation systems are vital as databanks grow.
• Understanding the relationship between structural and functional similarity is a key challenge.

Purpose of the Study:

• To quantitatively assess the consensus between CATH, SCOP, and FSSP protein domain annotation databases.
• To explore the graph-theoretical basis of annotation schemas (class, family, superfamily).
• To introduce and apply a measure of functional entropy to quantify structure-function relationships.

Main Methods:

• Generalized graph construction from CATH, SCOP, and FSSP databases.
• Hierarchical clustering of graphs and identification of overlap regions using distance measures.
• Development and application of a functional entropy measure to assess functional divergence and infer function from structure.

Main Results:

• Quantified the level of consensus between the three major protein domain annotation systems.
• Identified graph-theoretical phase transitions corresponding to annotation schema thresholds.
• Demonstrated a correlation between structural homology and functional proximity using functional entropy.

Conclusions:

• The study provides a rigorous quantitative analysis of consensus in protein domain annotation.
• Graph-theoretical properties underpin the observed annotation hierarchies.
• Functional entropy offers a novel way to measure and understand the limits of inferring protein function from structure.