The procedure of reference gene expression profiles construction and similarity score calculation. (1) Gene expression data annotated as treatment instances (i.e. treatment versus control) were extracted from GEO. (2) For each sample, genes were ranked in descending order according to the log ratio of the treatment to control. (3) Varying gene identifiers (gene names/IDs) were converted to UniGene IDs according to the associated platform annotation file. (4) Constructed rank vector of up- and down-regulated genes that matched between the query and reference, respectively, and sort the components in ascending order. (5) Calculated similarity score.