Fig. 6: Variation in the cohort target set (CTS) size and composition as function of the magnitude of filtering of genes expressed in non-cancerous human tissues, for different tumor types.

A, B The number of times a gene (cell-surface receptor) is included in the CTS (out of 20 replicates, which is therefore the Max count in panels A, B), where each column presents the CTS solutions when the input target genes sets are filtered using a specific TPM filtering threshold (“Methods”), for (A) breast cancer and (B) brain cancer. These data sets were selected due to their relatively small cohort target set sizes, permitting their visualization. C–F Circos plots of the genes occurring most frequently in optimal CTS solutions (length of arc along the circumference) and their pairwise co-occurrence (thickness of the connecting edge) for the four main cancer types, in our original target space of 1269 encoding cell-surface receptors. For each data set, we sampled up to 50 optimal CTS solutions. Co-occurrence representations of the 12 most common target genes out of 1269 encoding cell-surface receptors (with >5% frequency of occurrence) are represented in a cancer-specific manner for C brain cancer, (D) head and neck cancer (only seven genes have a frequency of 5% or more across optimal solutions), (E) melanoma, and (F) lung cancer. Genes and connections have distinct colors for improved visibility. Source data are provided as a Source data file.