Functional grouping of yeast genes via biclustering microarray data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray experiments generate large gene expression datasets.
• Standard clustering methods group genes across all conditions, potentially missing condition-specific patterns.
• Biclustering offers an alternative by identifying genes with similar expression across a subset of conditions.

Purpose of the Study:

• To investigate the feasibility and validity of a biclustering algorithm employing Gibbs sampling for analyzing yeast genome microarray data.
• To discover gene patterns and assess the effectiveness of biclustering compared to standard clustering.
• To explore the potential for deducing functions of unknown open reading frames (ORFs).

Main Methods:

• Developed a database for storing microarray and MIPS data.
• Implemented a biclustering algorithm utilizing a Gibbs sampling strategy in Matlab.
• Compared biclustering results with known gene annotations from MIPS and BLAST search similarity.

Main Results:

• The biclustering algorithm successfully identified gene patterns within the yeast microarray dataset.
• It demonstrated advantages over standard clustering by grouping genes over specific conditions with sharp probability distributions.
• The approach showed potential for inferring functions of previously uncharacterized ORFs.

Conclusions:

• The biclustering algorithm with Gibbs sampling is a viable method for analyzing gene expression data, offering transparent probabilistic interpretations.
• It provides more global and accurate gene patterns compared to expectation maximization methods, avoiding local minima.
• This approach aids in understanding gene function and discovering novel biological insights from microarray data.