Accurate Quantification of Gene Expression using Fuzzy Clustering Approaches

Despite the widespread application of microarray imaging for biomedical research, barriers still exist regarding its reliability and reproducibility for clinical use. A critical problem lies in accurate spot segmentation and quantification of gene expression level (mRNA) from microarray images. A variety of commercial and research freeware packages are available, but most cannot handle array spots with complex shapes such as donuts and scratches. Clustering approaches such as k-means and mixture models were introduced to overcome this difficulty, which used the hard labeling of each pixel. In this paper, we introduce a more sophisticated fuzzy clustering based method. We show that possiblistic c-means clustering performed the best among several fuzzy clustering approaches. In addition, we compared three statistical criteria in measuring gene expression levels and show that a new unbiased statistic is able to quantify the gene expression level more accurately. The proposed algorithms have been tested on a variety of simulated and real microarray images, demonstrating their better performance.