GDSC data

From release 1.0 of the Genomics of Drug Sensitivity in Cancer (GDSC) [22], we downloaded the following data files: gdsc_manova_input_w1.csv and gdsc_manova_output_w1.csv.

In gdsc_manova_input_w1.csv, there are 130 unique drugs as camptothecin was tested twice, drug ids 195 and 1003, and thus we only kept the instance that was more broadly tested (i.e. drug ID 1003 on 430 cell lines). Thus, effectively a panel of 130 drugs was screened against 638 cancer cell lines, leading to 47,748 IC₅₀ values (57.6% of all possible drug-cell pairs). Downloaded “IC₅₀” values are more precisely the natural logarithm of IC₅₀ in μM units (i.e. negative values represent drug responses more potent than 1 μM). We converted each of these values into their logarithm base 10 in μM units, which we denote as logIC₅₀ (e.g. logIC₅₀ = 1 corresponds to IC₅₀ = 10 μM), as in this way differences between two drug response values are directly given as orders of magnitude in the molar scale.

gdsc_manova_input_w1.csv also contains genetic mutation data for 68 cancer genes, which were selected as the most frequently mutated cancer genes [9], characterising each of the 638 cell lines. For each gene-cell pair, a ‘x::y’ description was provided by the GDSC, where ‘x’ identifies a coding variant and ‘y’ indicates copy number information from SNP6.0 data. As in Garnett et al. [9], a gene for which a mutation is not detected is considered to be wild-type (wt). A gene mutation is annotated if: a) a protein sequence variant is detected (x ≠ {wt,na}) or b) a deletion/amplification is detected. The latter corresponds to a copy number (cn) variation different from the wt value of y = 0 < cn < 8. Furthermore, three translocations were considered (BCR_ABL, MLL_AFF1 and EWS_FLI1). For each of these gene fusions, cell lines are identified as fusion not-detected or the identified fusion is given (i.e. wt or mutated with respect to the gene fusion, respectively). The microsatellite instability (msi) status of each cell line was also determined. Full details can be found in the original publication [9].