Figure 4.

Overview of connectivity mapping for estimating chemical concentration-dependent scores for a signature. (a) The query is a non-directional gene signature ($S$) signified by a set of genes (shown as black circles). (b) The reference is a transcriptomic profile $\mathit{x}$ shown as a vector of log2 transformed fold-change (L2FC) values for each gene (blue and red colors represent down- and up-regulation, respectively). (c) The similarity measure ($SM$) for scoring the match between $S$ and $\mathit{x}$. (d) A transcriptomic database (${\mathit{R}}_{\mathit{x}}$) comprised of a collection of $\mathit{x}$ for multiple chemicals and concentrations (Conc). ${\mathit{R}}_{\mathit{x}}$ is visualized as a matrix in which the rows represent genes, the columns show chemical concentrations, and the values in each column are $\mathit{x}$. For example, the outlined box in the matrix signifies eight $\mathit{x}$ for each of the concentrations of a chemical. (e) Concentration-response analysis of similarity scores between $S$ and $\mathit{x}(SM\left(S,\mathit{x}\right)$) for each $\mathit{x}$ of chemical shown in (d). The ordinate and abscissa show the similarity scores and the concentrations of the chemical, respectively. A null distribution (Null Dist.) of similarity scores (shown on the right of the graph along the ordinate axis) is generated by permuting ${\mathit{R}}_{\mathit{x}}$ and calculating $SM\left(S,\mathit{x}\right)$ for all random profiles. The standardized similarity scores (Z) (calculated using the null distribution and shown as “+” symbols) are analyzed by curve-fitting. The fitted concentration-response curve (blue) is used to estimate the benchmark concentration (BMC) corresponding to the benchmark response (BMR) value of Z=1.