The united signature algorithm (USA). This algorithm is designed to find a subset of conditions in which the input genes are regulating each other or are co-regulated, and to identify additional genes that are potentially co-regulated under the same subset of conditions. The order of the procedures performed in the USA is shown in the following six panels: a) bi-normalization and log transformation of the raw expression data, such that row sums and column sums are equal to zero, b) selection of an input set of gene expression profiles consisting of the target gene and its TF regulator, or the expression profiles of the target gene and all the other known regulated genes controlled by the TF. c) calculation of condition (column) scores by summing (or averaging) the columns of a sub-matrix, whose rows represent the normalized expression profiles of the input genes across all conditions. These rows are first multiplied by +1 for input genes that are stimulated by the TF and by -1 (inversion) for target genes inhibited by the TF. Experimental conditions whose column average Sc across the input genes satisfies |Sc-mean(Sc)| > thresholdcolumn are retained as indicated by black bullets and black experimental IDs below the sub-matrix, d) calculation of gene (row) scores defined as the weighted row average S
)/(#genes) across the selected conditions c e) determination of a sub-matrix of genes and conditions, termed the united transcriptional module (UTM), consisting of gene expression profiles whose weighted row averages satisfy |Sg-mean(Sg)| > thresholdrow across the selected conditions c, f) retaining genes within the UTM whose correlation with the target gene Gene2, TF or the centroid gene (a gene that is correlated with the largest number of genes within the UTM) satisfy |R| > α = thresholdcorrelation where R is the correlation coefficient.