To extend this analysis to include specialized primary cell types, we considered 768 compounds that had been profiled in both neural progenitor cells (NPC) and differentiated neurons (NEU) as well as the 9 core cancer cell lines. For each compound, we computed the similarity, using WTCS, between all pairwise combinations of cell lines and converted to τ using the pairwise similarities between all 768 compounds in all 11 cell lines as reference (Qref). We observed that 189 of the 768 compounds (25%) connected with τ ≥ 90 when comparing NPC to NEU. For each pairwise combination of the 11 cell lines (NPC, NEU + 9 core) we computed the fraction of the 189 compounds that self-connected above 90. We then computed the average fraction that self-connected when considering NPC to cancer (34%), NEU to cancer (25%) and cancer to cancer (50%). This suggests that the neuronal lines are more different from the cancer lines than the cancer lines are from each other, at least in the space of these 189 compounds. Therefore, expanding the cell line set into neuronal cell types may be beneficial.
