Fig. 7: Association of somatic mutations with drug response. | Nature Communications

Fig. 7: Association of somatic mutations with drug response.

From: Deep generative neural network for accurate drug response imputation

Fig. 7

A Volcano plot of drug–gene association. The genomic signature was defined using the mutation status of each gene. Each dot represents the statistics of the association status of a gene in a cancer type with a drug. The P value was obtained from a two-sided Wilcoxon test comparing the predicted response in samples with the mutant gene (i.e., those with deleterious missense SNVs or nonsense SNVs) (MT) and samples with wild-type gene (WT). The former group was required to have at least ten samples. x axis: the difference of the average drug response in the samples harboring the mutated gene from that in the wild samples, y axis: −log10(P) where the unadjusted p-value was used for plotting. Red (difference >0) and blue (difference <0) dots indicate significant associations (BH-adjusted P < 0.05). B Heatmap showing the significant drug–gene associations in each cancer type. Due to space limitation, only associations with adjusted PBH < 0.005 for positive association or PBH < 0.01 for negative association were plotted. Associations involving the gene TP53 were also excluded from plotting. The color and size of the circle in each cell is proportional to -log10(p) (note that we used the unadjusted P value for plotting). C Heatmap showing the significant drug–gene associations determined by mutation clusters. Each cell represents a mutation cluster of a specific gene in association with a specific drug. The P value was obtained from a two-sided Wilcoxon test comparing the predicted response in samples with mutations in a mutation cluster (i.e., only deleterious missense SNVs or nonsense SNVs considered) and samples with wild-type gene. Similarly, the former group was required to have at least ten samples to allow more tested clusters. D Distribution of the response to drug AZD6244 in the samples with two mutation clusters of Ras genes. Only the cases that had five or more mutated samples are shown. Each box shows the interquartile range (IQR between Q1 and Q3) for the corresponding set. The central mark (horizontal line) shows the median. E Association of drug response with copy number gain. F Association of drug response with copy number gain, focusing on chromosome 17 where ERBB2 was located.

Back to article page