• 2022-09
  • 2022-08
  • 2022-07
  • 2022-05
  • 2022-04
  • 2021-03
  • 2020-08
  • 2020-07
  • 2018-07
  • br TSG resulting in transcriptome wide t values


    TSG, resulting in transcriptome-wide t-values for all these genes. qffiffiffiffiffiffiffiffiffiffiffiffi
    We defined a combined impact score T = t21 + t22, where t1 is the t-value for the TSG itself (cis-effect) and t2 is the average t-value for the top 1% differentially expressed genes (DEGs) in the transcriptome excluding the TSG itself (trans-effect).
    Further filtering was performed to remove abnormal outliers. As a result, 208/277 (75.1%) events remained unchanged, 53 (19.1%) were reduced by less than 4 samples, 5 (1.8%) were reduced by 4–7 samples, and 11 events were excluded due to insufficient number of samples (Figure S7). Hereafter, we use inactivated samples to refer to the samples from L2+L1 after this filtering process.
    To explore the distribution of cis- and trans-effects, we sepa-rated the remaining 266 TSG 3 cancer events into four sub-groups according to whether their cis-effect was nominally significant (pcis < 0.05, corresponding to t1 in DEG analysis) and whether their trans-effect was significant (ztrans > 1.28, cor-responding to FDR < 0.20, where ztrans was calculated as the average z-score transferred from pBH from DEG analysis) (Fig-ure 3D). First, the most common pattern included TSG 3 cancer events with both a cis-effect and a trans-effect (n = 150). This finding is expected and supports that genetic inactivation events lead to decreased SP 600125 of TSGs (cis-effect), which in turn regulates their downstream genes (trans-effect). Muta-tions in these events are more likely to contribute to cancer. Second, 39 TSG 3 cancer events had no cis-effect but had a substantial trans-effect (blue dots in Figure 3D). We illustrate this using two cases: MAP3K1 (Figure 3B) and GATA3 (Fig-ure 3C). MAP3K1 had inactivation mutations in 44 BRCA_LumA samples but no sign of decreased expression: all with trunca-tion mutations (L1). However, broad DEGs were observed in these 44 samples compared with MAP3K1 WT samples. A similar trend was observed for GATA3 in both BRCA_LumA (51 inactivated samples with truncation mutations (L1), not shown in the figure) and BRCA_LumB (29 inactivated samples with truncation mutations, L1). Third, 53 TSG 3 cancer events had a cis-effect, but not much of a trans-effect (green dots in Figure 3D). Examples included KDM6A in BLCA, ESCA, HNSC, LUSC, and PAAD; ARID1A in BLCA, LIHC, and LUAD; and NCOR1 in LIHC and STAD. Finally, there were 24 TSG 3 cancer events with neither a cis-effect nor a trans-effect (gray dots in Figure 3D). The specific molecular mechanisms of these events on the transcriptome level require further inves-tigation, with possible roles, such as abnormal transcription or splicing. In summary, our analysis revealed that 91.0% TSG 3 cancer events had effect in the forms of a cis-effect, a trans-effect, or both.
    TSG Inactivation Events: Global versus Focal Impact
    weak trans-effect. The variation in DEGs was unlikely due to the small number of inactivated samples, because a similar trend of DEGs was still observed when requiring nmut R 20 for each TSG 3 cancer event (Figure 3G). These results suggested that the strong variation of DEGs was likely due to TSG’s biological disturbance rather than random sampling.
    To evaluate the significance of the observed transcriptomic footprints of these TSG 3 cancer events, we implemented a randomization test and found that in 161 (60.5%) TSG 3 cancer events (71 unique TSGs), the TSGs had significantly stronger transcriptomic impact than what would be randomly expected (#DEGs R 10, pemp < 0.05) (Figure 3F).
    Strong TSG Inactivation Characteristics beyond Cancer Lineages
    Next, we explored whether inactivation of the same TSG would lead to similar transcriptional footprints in different cancer types. We used transcriptome-wide t-values of each TSG 3 cancer event to perform a hierarchical cluster analysis and particularly focused on the 161 TSG 3 cancer events with a sig-nificant impact (#DEGs R 10, pemp < 0.05). An initial application illustrated two distinctive groups, where RB1, TP53, and CDKN2A were frequently observed to cluster together but within-cancer groups were also observed (Figure S7). To con-trol for potential confounding factors of shared samples (e.g., one sample with inactivated events by two TSGs), we restricted the analyses only to those samples that did not share mutant TSGs. Specifically, when comparing two TSGs, e.g., TSG_1 (in-activated samples denoted by MT_1 and wild-type samples by WT_1) and TSG_2 (MT_2 and WT_2), we chose to use only MT_1 and WT_1 in the wild-type samples of TSG_2 (WT_2), and vice versa. As a result, for any pair of TSGs compared, the transcriptome-wide t-values for a TSG were only obtained in the wild-type of the other TSG, and, hence, the potential con-founding impact by the other TSG was controlled. Using the transcriptome-wide t-values for each TSG 3 cancer event, we conducted hierarchical clustering and observed two major clusters (Figure 4A). TSGs involved in cell-cycle regulation, such as TP53, RB1, CDKN2A, and PTEN, were overrepresented in the larger group. Although some cancer types were occa-sionally enriched in local subclusters (e.g., LGG and KIRC), the overall cluster showed enrichment independent of cancer types, suggesting common pathways might be disturbed upon the inactivation of the same TSGs in different cancer types. In addition to the RB1, PTEN, CDKN2A, and TP53 clus-ters, we also observed local structures, such as STK11, FBXW7, BAP1, and NCOR1.