<![CDATA[This study proposes a three-phase biclustering approach to detect strong correlations between genes and conditions in Diabetes Mellitus (DM) gene expression data, focusing on obese and lean samples. The first phase employs Singular Value Decomposition (SVD) to transform the data into gene- and condition-based matrices. The second phase applies Partition Around Medoids (PAM) with Euclidean distance to generate initial biclusters. The third phase evaluates biclusters using a modified Pearson correlation-based lift algorithm, optimized for detecting additive-multiplicative patterns. Implementation on microarray data produced δ-corbiclusters with high gene-sample correlations. Applied to a DM dataset (1,331 selected genes), the SVD-PAM phase yielded 8 initial biclusters, while the modified lift algorithm refined these into 3 δ-corbiclusters with correlation values of 0.097, 0.095, and 0.085, respectively. These results demonstrate the method’s effectiveness in revealing significant gene-condition relationships in DM, highlighting its potential for advancing medical research on diabetes.]]>
Copyrights © 2026
