Figure 1

We search for time scale separation in microarray expression data by detecting patterns that show a temporal shift between experimental replicates relative to the synchronization by a given biological process. In this example, Gene A is related to the synchronization procedure and shows very similar expression patterns in both replicates. The expression level of Gene B on the other hand is also changing in a reproducible way, but on a different time scale, independent of the synchronization process. By shifting these time series relative to each other and calculating the similarity for each possible shift, we find the superposition that yields the maximal similarity in the overlap window (third column). Because of the varying window size we keep track of the statistical significance (see Methods). The value of the shift s which gives rise to the maximal similarity, as well as the statistical significance of this value, allows us to determine whether a given expression pattern is likely to be evidence of a process operating on a separate biological time scale.