Figure 1

The flowchart of the approach. The pre-processing section refers to filtering for differentially expressed probesets in each dataset, mapping to gene symbols to extract a set of common genes that are present across all datasets, and then re-mapping to corresponding probesets in each particular dataset. The main steps include establishing the AM to characterize how much confidence two probesets (and two genes) are co-expressed in each condition (and then across all conditions) and searching for all possible clusters of co-expressed genes based on the common AM. The post-processing step will select those clusters that are significant and optionally merge those with similar expression patterns if indicated.