Fig. 1.
(A) Frequently observed patterns of correlation changes revealed by rough clustering (rmsd <0.1), (B) schematic explanation of stability values, S, for the correlation change, (C) the distribution of stability values for all pairs we consider (black), the pairs with decrease cases (green) and those with increase cases (red), where the number of probe pairs, the mean of the S values and its standard deviation for the decrease cases are 430 866 (76.5%), 0.533 and 0.159, respectively, while those for the increase cases are 132 719 (23.5%), 0.547 and 0.199, and (D) frequency plot to show the relationship between correlation and stability. Frequency is shown in 10 base logarithmic scales. The stability tends to be high for the gene pairs with >0.7 correlations, but for the pairs with <0.7 correlations their correlations are usually fragile as indicated by low stability.