K-means clustering groups objects based on their similarities, where k is the number of clusters. A dynamic second-order k-means (DSOKM) clustering is characterized by i) k acting as an independent variable, and ii) the transformation of original data into a secondary form before cluster analysis (e.g., mapping original data on to the elastic principal grid (EPG) of A. Gorban and A. Zinovyev). We employed a DSOKM clustering method implemented with ViDaExpert to analyze the kinetic behaviors of the genome-wide mRNA levels measured in budding yeast by Garcia-Martinez al. [Mol. Cell 15:303-313 (2004)]. The glycolytic and oxphos transcripts were found to cluster into two separate regions on EPG and exhibited distinct dynamics as k is varied from 4 to 84 in 6 steps. To quantify the kinetics of the cluster motions on EPG, we defined an order parameter called �transcript density (d_T)� as d_T = k(the % of the total transcripts with a given function localized within a set of contiguous clusters)/(the fraction of the total grid area occupied by contiguous clusters). When d_T values of various functional transcripts are plotted against k on a log-log scale, straight lines were obtained obeying the power law, d_T = ak^w, where a is a proportionality constant, and w is the 'critical' exponent which varied with the biological functions of clusters. We found that protein synthesis transcripts and transcripts with unknown functions had similar w values (0.29 vs. 0.28) as compared to others (0.65-0.78), suggesting that most of the 290 unknown RNA molecules may participate in protein synthesis as protein-synthesis-regulators [J. S. Mattick, Nature Rev. Genetics 5:316-323]. Thus the DSOKM clustering method has demonstrated its utility in classifying intracellular dissipative structures [S. Ji, J. theoret. Biol. 116:399-4126 (1985)] and may prove equally useful in studying other more complex dissipative structures, including EEG and EKG.