Dynamically weighted clustering with noise set.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene expression data analysis often employs clustering methods to identify co-expressed genes.
• Functional annotation data, such as Gene Ontology (GO), offers potential to enhance clustering performance.
• Identifying genes with distinct expression profiles (scattered genes) can improve clustering accuracy.

Purpose of the Study:

• To develop a novel clustering algorithm that integrates gene annotation information.
• To enable the exclusion of scattered genes (noise set) from main clusters.
• To improve the functional consistency and expression coherence of identified gene clusters.

Main Methods:

• Developed Dynamically Weighted Clustering with Noise set (DWCN) algorithm.
• Utilized gene annotation data within the clustering process.
• Tested DWCN against common clustering techniques on simulated and public yeast gene expression datasets.

Main Results:

• DWCN successfully identified scattered genes, forming a 'noise set'.
• The algorithm demonstrated improved performance compared to existing clustering methods.
• Analysis included simulated data, Stanford yeast cell-cycle data, and yeast segregant data.

Conclusions:

• DWCN generates clusters with superior functional annotation consistency.
• The method produces more coherent gene expression patterns than traditional clustering techniques.
• DWCN offers an effective approach for analyzing gene expression data with functional annotations.