Age-related feature extraction on mouse skeletal muscle: Data mining approach

We applied a rigorous data mining strategy that combined principal component analysis and a clustering algorithm to analyze twenty-eight structural and compositional characteristics of mouse skeletal muscle at four different age groups: 2, 11, 22 and 25 months. First, principal component analysis was implemented to (1) optimize the data by selecting two out of twenty-eight dimensions to represent the primary age-related information; (2) derive information about the extent of the correlations between the individual characteristics and ages. It is revealed that the structure of extracellular matrix weakly correlated to aging whereas the morphology of muscle cells and distribution of nuclei presented strong correlations. Next, several hierarchical clustering algorithms were applied within the selected two-dimensional space to evaluate the differences among the age groups. Ward's aggregative method was shown to perform the best. The cluster structure derived from Ward's method reveals a significant difference between mice at 25 months and the other age groups. However, differences in the structural characteristics are marginally significant between mice at 22 months and 2-11 months.