Local Clustering

Local Clustering identifies time-delayed and inverted relationships in gene expression data to detect activation and inhibition interactions and reveal localized regulatory relationships overlooked by global correlation methods.

Key Features:

• Local clustering methodology: Identifies relationships based on temporal shifts or inversions in expression profiles, analogous to local sequence alignment (Smith-Waterman) extending principles of global alignment (Needleman-Wunsch).
• Statistical significance: Evaluates each identified cluster using random score distributions to assess significance against chance.
• Application to gene-expression datasets: Applied to yeast cell-cycle expression datasets to reveal additional biological relationships beyond traditional correlation methods.

Scientific Applications:

• Functional annotation: Uses time-delayed or inverted expression profiles to infer functions of uncharacterized genes and predict shared protein-protein interactions or similar cellular roles.
• Biological insight: Corroborates activation and inhibition examples such as the YME1–YNT20 relationship in yeast and uncovers new relationships involving uncharacterized genes.
• Broader dataset utility: Applicable to datasets beyond yeast cell-cycle data to identify localized temporal regulatory relationships.

Methodology:

Detects localized temporal relationships by identifying time-delayed and inverted expression patterns and assesses cluster significance using random score distributions; conceptually applies local alignment principles (Smith-Waterman versus Needleman-Wunsch) to gene-expression analysis.