Pattern Recognition

If more complex questions are to be asked of a microarray experiment than merely identifying which genes are up- or downregulated, more complex procedures need to be performed with the data. For example, researchers often use cluster analysis to group genes with similar expression profiles. A cluster algorithm iterates over an array data set and acts as a grouping tool, essentially assigning two genes with similar expression profiles to the same cluster. There are many types of cluster algorithm including hierarchical, k-means, principal component analysis (PCA), self-organizing maps (SOMs), etc. In this way clustering can help in the assignment of functionality to unknown genes, divulge groups of genes controlled by the same promoter, and generally provide a more holistic integration of microarray data. However, as with all statistical analysis, clustering can never be perfect, primarily because it requires user input to determine cluster cutoff values. For more detailed descriptions of clustering and its uses, see Ref. [43].

0 0

Post a comment