Fig 1.
Power comparison of mSAME (red bars) and GLM (blue bars) for mutation-level simulations.
Fig 2.
Power comparison of gSAME (red bars) and GLM (blue bars) for gene-level simulations.
Table 1.
Nine significant results detected uniquely by mSAME.
An eQTL is a local eQTL if the distance between the somatic mutation and the gene is smaller than 1Mb. Otherwise it is a distant eQTL.
Table 2.
Summary for the association study of the subtypes on mutation level (top table) and gene level (bottom table).
Fig 3.
Summary of pan-cancer eQTL mapping results by SAME with Bonferroni multiple testing correction.
Left panel: a heatmap of mutation-level eQTL mapping results by mSAME across cancer types. Each cell in the heatmap is colored according to the number of significant associations for one mutation (row) in one cancer type (column). Only those mutations that are associated with 2 or more genes across the 12 cancer types are shown. Note that OV cancer is not included since there is no significant eQTL in OV. Right panel: a heatmap of gene-level eQTL mapping results by gSAME across cancer types. Only those gene-level mutations that are associated with 6 or more genes across the 12 cancer types are shown.
Fig 4.
Summary of eGenes of TP53 with transcriptome-wide multiple testing correction.
The results of gSAME (left panel) and GLM (right panel) were summarized by heatmaps showing whether a gene is eGene across cancer types. Only the genes that are eGenes for three or more cancer types by either method are shown. The asterisk (*) next to gene symbol indicates a gene is a high confidence TP53 target gene [37].
Fig 5.
Comparison of eQTL mapping using mSAME p-values versus gSAME p-values on -log10 scale in colon cancer patients.
Left panel: mutation-level versus gene-level analysis for the oncogene BRAF for all gene expression traits with eQTL p-value smaller than 0.05/16, 339 by either mSAME or gSAME. Right panel: similar results for the tumor suppressor gene TP53. Dashed line indicates y = x.