Skip to main content
Figure 1 | BMC Systems Biology

Figure 1

From: New network topology approaches reveal differential correlation patterns in breast cancer

Figure 1

Workflow of the algorithms for detection of differentially correlated genes. In the first step, the gene correlation matrix is calculated for each of the disease conditions. In the second step, correlation networks are constructed for a fixed correlation threshold. Two genes are connected with an edge whenever the Pearson correlation exceeds this threshold. The differences in local (algorithm DCloc) or global (algorithm DCglob) topology of the networks are analyzed. Step two is repeated for a series of thresholds (typically 100) such that a good coverage of the correlations in the data set is obtained. The series of thresholds is chosen as equidistant sequence of Fisher-transformed correlations. In the third step, the results for the thresholds are averaged and a measure of differential correlation is calculated for each of the genes. After choosing a cutoff point for the measure of differential correlation, a list of genes with higher correlation in condition A and a list of genes with higher correlation in condition B are obtained.

Back to article page