br various kinds of mutations as
various kinds of mutations as shown with three or more colorful stacks in Figure 2D. Although deleterious missense mutations were included, they were found mainly in a few genes, such as TP53 (multiple cancer types), PTEN (GBM), KEAP1 (LUAD),
and VHL (KIRC). Most genes were found with both L2 and L1 mu-tations, as shown with four colorful stacks in Figure 2D. Many genes with deep SB 431542 were filtered out by the sample-based test because they mainly, and sometimes only, occurred in
Figure 3. Inactivation Events of TSGs with Cis- and Trans-Effects
(A) Evaluation of TSG 3 cancer inactivation events using transcriptomic data. In each panel, the group labeled ‘‘high’’ indicates the significant TSG 3 cancer events (psample < 0.05 and pgene < 0.05), and the group ‘‘low’’ indicates non-significant events. Cis- and trans-effects were assessed for L1 mutations and L2 mutations, respectively.
(B and C) MAP3K1 (B) and GATA3 (C) as example genes that had no cis-effect but had a trans-effect. In both cases, as shown, although MAP3K1 and GATA3 were both found with inactivation mutations, their mRNA expression level was not decreased compared to the WT samples of the corresponding genes.
(D) Plot of all TSGs for their cis- and trans-effects. Each dot indicates a TSG 3 cancer event. Node color indicates the events with different cis- or trans-effects. Nodes with a black circle indicate the 161 events with significant impact.
(E) Distribution of DEGs (y axis) versus T (see main text for the definition of T). Each dot represents a TSG 3 cancer event. Node color indicates cancer types (like in F), and node size is proportional to the number of inactivated samples.
(F) Distribution of differentially expressed genes (DEGs, y axis) versus empirical pemp (x axis) of each TSG in each cancer type. Color for cancer types are the same as those in (E).
(G) Distribution of significant DEGs.
samples with a high mutation load. Hence, they were likely pas-sengers resulting from genome-wide accumulation of mutations.
To assess the biological significance of TSG 3 cancer inacti-vation events, we compared the cis- and the trans-effect be-tween significant and non-significant TSG 3 cancer events using TCGA transcriptome data (Figure 3A). Sixty cases had sufficient data (R5 testable TSGs and R3 significant TSGs). Among them, the significant TSGs in 48 cases had a stronger cis-effect and a trans-effect than did non-significant TSGs (Figure S6). This indi-cated that the TSG 3 cancer inactivation events identified by our methods at the DNA level could be supported by their gene expression.
When two or more L2 TSGs are located in a region overlapped with a CNV deletion, we clustered these TSGs and selected one to represent each cluster. For example, in the CDKN2A-CDKN2B-MTAP cluster, we selected CDKN2A for the following analysis. In this process, we obtained 277 TSG 3 cancer events for tran-scriptomic analyses, representing unique inactivation events.
TSG Inactivation Events at the Transcription Level
Next, we explored the impact of genetic inactivation events on the TSGs themselves (cis-effect) and on other genes (trans-
effect) using the transcriptome data. For the cis-effect, although we expected that nonsense SNVs and copy-number loss would result in decreased expression, this was not always the case in the actual data because of various reasons, such as unusual bio-logical mechanisms (e.g., escaping NMD or stop codon read through) (Holbrook et al., 2004), technique issues (e.g., sequencing errors and sample purity), and variant calling accu-racy. In this work, a trans-effect is defined as a TSG 3 cancer event that influences the expression of other genes, such as downstream genes in the signaling pathways. It is commonly recognized that the passenger mutations irrelevant to cancer development would not have much impact on other genes, whereas driver mutations are expected to interrupt signaling pathways that are critical to cancer. To measure both cis- and trans-effects, a differential expression analysis for each TSG 3 cancer event was performed by comparing the expression pro-files of L2+L1 samples with the matched WT samples for each