Archives

  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • 2021-03
  • br Validation of selected top associations between

    2020-08-18


    Validation of selected top associations between gene Erastin and drug response in the NCI-60 dataset
    After observing statistically significant associations (Table 1) between OCM gene expression and drug sensitivity in the 635 cell lines with available CCLE and GDSC measurements, we validated several strongest associations using the data from the NCI-60 cancer cell line panel. The 60 cell lines from that panel have been extensively characterized for their molecular features and screened manually for drug response at the U.S. National Cancer Institute (NCI) [35]. Microarray gene expression data (in the log2 format, representing an average of measures from five microarray platforms includ-ing Affymetrix HG-U95, HG-U133, HG-U133 Plus 2.0, and GeneChip Human Exon 1.0 ST arrays, and Agilent Whole Hu-man Genome Oligo Microarray) were downloaded using the CellminerCDB resource v. 1.0 at the National Cancer Institute [24,36].
    We identified the top 20 strongest associations in the CCLE-GDSC dataset with the highest values of Pearson r listed in Table 1. These correlations involved 16 agents. For five of these agents, crizotinib, dasatinib, erlotinib, lapatinib, and methotrexate, drug response values, which had been generated through the NCI-60 screening by the NCI Develop-mental Therapeutics Program, were available for download from CellminerCDB. These NCI-60 data were downloaded from CellminerCDB in the format of z-scores computed from -log10(GI50), where GI50 is a molar concentration resulting in
    a 50% growth inhibition [24,36]. These z-scores were multi-plied by −1 for comparability of the directions of associations with drug response between the GDSC-CCLE and NCI-60 datasets. We further refer to these transformed NCI-60 scores as the z-scores of log(GI50) values. Among the top 20 cor-relations listed in Table 1, GDSC sensitivity measures to 5 agents were associated with CCLE-measured expression of 6 OCM genes. For each of these 6 drug-gene pairs, we selected the top 10% and the bottom 10% of the NCI-60 cell lines with the highest and the lowest expression levels of the OCM gene involved in each correlation. This resulted in six cell lines with the highest and six cell lines with the lowest expression of the OCM gene for each specific correlation. We used the Stu-dent’s t-test to test whether the NCI-60 cell lines groups with high and low folate gene expression had statistically signifi-cantly different average drug response z-scores. The resulting p-values were FDR adjusted. The effect sizes for the compar-ison of drug response between the highest and the lowest 10% of the OCM gene-expressing cell lines were estimated using Cohen’s d, which was computed using the numpy pack-age v. 1.11.3 [37]. In addition to comparing the differences in drug sensitivity between the NCI-60 cell lines with high and
    Table 1 One-carbon metabolism genes which were associated with sensitivity to antitumor agents with r > 0.3 and FDR adjusted p < 0.05.
    Agent Targets Associated Pearson r FDR adjusted Sample Spearman Targeted pathway
    gene
    p-value size
    D.-J. Min,S. VuralandJ. Krushkal
    Table 1 (continued)
    Agent Targets Associated Pearson r FDR adjusted Sample Spearman Targeted pathway
    gene p-value size
    Other 595 ∗ Other
    (continued on next page)
    Table 1 (continued)
    Agent Targets Associated Pearson r FDR adjusted Sample Spearman Targeted pathway
    gene p-value size
    Other; kinases 595
    PI3K/MTOR signaling 595 ∗ PI3K/MTOR signaling 595 ∗ RTK signaling 595 ∗ RTK signaling 227 ∗ RTK signaling 227 ∗ RTK signaling 227 ∗ RTK signaling 227 ∗ RTK signaling 595
    RTK signaling 222 ∗ RTK signaling 222 ∗ RTK signaling 222 ∗ RTK signaling 630
    WNT signaling 
    Pearson r: Pearson correlation coefficient; Spearman: an asterisk (∗ ) indicates correlations which also satisfied ρ > 0.3 and FDR adjusted p < 0.05 for Spearman correlation.
    Information about molecular drug targets and target pathways for each agent listed in Table 1 was obtained from the GDSC data download site [25]. Additional information about ABL signaling inhibitors was included based on data from Supplementary Table S1F of Iorio et al. [27]. An expanded list of potential drug targets for each agent is provided in Supplementary Table 1.  VuralandJ. Krushkal
    low OCM gene expression, we also used Pearson correlation analysis to evaluate the associations between OCM gene ex-pression and drug response for these 6 drug-gene pairs in the entire NCI-60 dataset.
    Analysis of correlation of expression among OCM genes, and of OCM gene expression with regulatory genes and genes involved in drug target pathways