Visualizing and Clustering Protein Similarity Networks: Sequences, Structures, and Functions.

Area of Science:

• Proteomics
• Bioinformatics
• Molecular Evolution

Background:

• Protein network knowledge is crucial for understanding molecular evolution, cellular robustness, and protein function annotation.
• Existing methods may present inconsistencies when classifying proteins based on sequence, structure, and function.

Purpose of the Study:

• To develop a general clustering approach for visualizing the sequence-structure-function relationships within protein networks.
• To investigate the reasons behind inconsistencies in protein classification based on sequence, structure, and function.
• To facilitate a better understanding of protein relationships and aid in comprehending protein databases.

Main Methods:

• Applied the minimum span clustering (MSC) method to cluster 1437 enzymes based on their sequences and structures.
• Delineated the protein network structure at two clustering resolutions.
• Calculated Jaccard's similarity coefficients to quantify the consistency between sequence, structure, and function classifications for proteases.

Main Results:

• The second-level MSC clustering demonstrated high similarity to established enzyme classifications.
• Clustering results showed consistency across sequence, structure, and function data for enzymes.
• High Jaccard's similarity coefficients (0.86 for sequence-function, 0.82 for sequence-structure, 0.78 for structure-function) were observed for proteases.

Conclusions:

• The developed clustering approach provides a comprehensive view of the protein sequence-structure-function network.
• This visualization aids in intuitively understanding relationships between proteins and predicting the structure and function of new protein sequences.
• The findings support the utility of integrated sequence, structure, and function data for robust protein classification and evolutionary studies.