Clustering protein environments for function prediction: finding PROSITE motifs in 3D.

Area of Science:

• Computational biology
• Structural biology
• Bioinformatics

Background:

• Structural genomics yields many protein structures lacking functional annotation.
• Supervised methods like FEATURE are limited to known functional sites.
• Many functional protein sites may remain undiscovered.

Purpose of the Study:

• To develop a method for clustering protein microenvironments.
• To identify novel, previously unrecognized functional sites in proteins.
• To evaluate the potential for discovering new functional roles based on structural similarity.

Main Methods:

• Developed a computational method for rapid clustering of protein microenvironments.
• Utilized a simplified representation and K-means clustering algorithm.
• Clustered approximately 2,000,000 microenvironments from 9,600 protein chains, defining 4,550 clusters.

Main Results:

• Successfully clustered millions of protein microenvironments.
• Identified 4,550 distinct clusters representing similar residue environments.
• Validated the method by rediscovering known 3D environments associated with PROSITE motifs, finding clusters enriched for shared motifs.

Conclusions:

• Protein environments can be effectively clustered using simplified representations and K-means.
• Rediscovery of known motifs validates the clustering approach and aids in calibrating cluster parameters.
• This method facilitates the discovery of novel structural and functional sites by identifying new, similar clusters.