Schematic of the template match method. Hypothetical templates and genes are depicted as graphs of arrays (horizontal axis) vs expression (arbitrary units, vertical axis). The replicates for each condition are listed as 1 and 2 (for example, 129SvEv - Ag1 and 129SvEv - Ag2 are the two amygdala samples from 129SvEv). Two example templates are depicted schematically, and represent the profiles of idealized genes showing the patterns of interest. The top template is an idealized profile designed to select for genes showing midbrain enrichment or depletion, and the lower one is designed to select for genes showing strain differences across all regions. Each gene in the dataset is then ranked by its correlation with the template. In the hypothetical examples shown, gene 1 most closely resembles the strain template, with gene 2 giving the next best match. Gene 4 is a fairly weak match to the midbrain template. Gene 5 is a strong match to the reverse of the midbrain template; the use of the absolute value of the correlation coefficient allows the midbrain template to select both genes 4 and 5. Gene 3 resembled neither template shown here. In the actual implementation, all 13,067 genes are compared to a given template and ranked by the quality of the match. Ag, amygdala; Cb, cerebellum; Cx, cortex; Ec, entorhinal cortex; Hp, hippocampus; Mb, midbrain.
Pavlidis and Noble Genome Biology 2001 2:research0042.1 doi:10.1186/gb-2001-2-10-research0042