Validating clustering for gene expression data
Ji and Tan  focus on extracting time-lagged gene clusters known as q-clusters, where is the time length of a bicluster (i.e., the number of consecutive time points in the bicluster), that can have different time lengths, but genes in the same cluster must have the same durations over time, even though time lags exist among the genes.  proposed to use a wavelet-based cluster method to detect time shift/delay situation.
They then rely on string processing techniques to develop an algorithm that identifies contiguous column coherent biclusters.  alter original expression data by deleting and inserting border time points, and then use an algorithm based on a mean squared residue score to cluster the modified expression data.
We compare different levels of genome-wide co-clustering by weighting the involved sources of information differently.
Clustering quality is determined by both general and SOM-specific validation measures.
We know that standard exploratory clustering methods are useful for grouping items that behave in a similar fashion.
However, when these standard approaches are applied to experiments that evaluate the transcriptome over coordinated experimental stages, they fail to acknowledge the dynamic nature of such processes.