Possibilistic reformed fuzzy local information clustering technique for noisy microarray image spots segmentation

The cDNA microarray image provides useful information about thousands of gene expressions simultaneously. This information is used by bioinformatics researchers for diagnosis of different diseases and drug designs. Microarray image spot segmentation using an improved fuzzy clustering algorithm is proposed in this article. The proposed Possibilistic Reformed Fuzzy Local Information C Means (PRFLICM) algorithm is a variant of Possibilistic Fuzzy Local Information C Means (PFLICM) algorithm. The parameters used for testing the proposed algorithm include segmentation matching factor (SMF), probability of error (p(e)), discrepancy distance (D), normalized mean square error and sum of square distance (SSD). The performance of the algorithm is validated with a set of simulated cDNA microarray images with known gene expression values. From the results of SMF, the proposed PRFLICM shows an improvement of 0.4% and 0.1% for high noise and low noise microarray images respectively when compared to PFLICM algorithm. The proposed algorithm is applied to yeast microarray database (YMD) and is used to find the yeast cell life cycle generated genes. The results show that the proposed algorithm has identified 101 cell life cycle regulated genes out of 104 such genes published in the YMD database.