Clustering approach to detect mRNA–degradation patterns from DNA–microarray gene-expression data

DNA-microarray based gene-expression analysis is based on hybridization events between messenger RNA (mRNA) and single stranded DNA probes. In oligo nucleotide DNA-microarrays the probes consist of approximately 20 to 80 nucleotides long DNA-molecules. Consequently, several unique probes perfectly matching each single open reading frame (ORF) of monoor polycistronic mRNA are usually used. If these probes are distributed over the whole length of the mRNA molecule, information about mRNA-degradation patterns can be gathered with data clustering methods. Here we report analysis of expression of 1107 open reading frames from the cyanobacterium Nostoc PCC 7120. Each open reading frame was covered by 10 unique 25 nucleotides long probes and analyzed by 4 independent DNA-microarray experiments. Both the positional information and the absolute expression value for each probe were used to infer clusters of transcripts that show similar expression patterns. Hierarchical and fuzzy k-means clustering yielded comparable results. Our results suggest that several different mRNA-degradation mechanisms, specific for certain transcripts, work in concert.