Leveraging methylation to identify the potential causal genes associated with survival in lung adenocarcinoma and lung squamous cell carcinoma

Abstract

Understanding the different genetic landscape between lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) is important for understanding the underlying molecular mechanism, which may facilitate the development of effective and precise treatments. Although previous studies have identified a number of differentially expressed genes (DEGs) responsible for lung cancer, it is unknown which of these genes are causal. The present study integrated DNA methylation, RNA sequencing, clinical characteristics and survival outcomes of patients with LUAD and LUSC from The Cancer Genome Atlas. DEGs were first identified using edgeR by comparing tumor and normal tissue, and differentially methylated probes (DMPs) were assessed using ChAMP. Candidate genes for further time-to-event instrumental variable analysis were selected as the intersecting genes between DEGs and the genes including DMP CpG sites within the transcription start site (TSS1500), with DMPs in TSS1500 region being the instrumental variables. Extensive sensitivity analyses were conducted to assess the robustness of the results. The present study identified 906 DEGs for LUAD, among which 538 also had DMPs in the TSS1500 region. In addition, 1,543 DEGs were identified for LUSC, among which 1,053 also had DMPs in the TSS1500 region. Time-to-event instrumental variable analysis detected eight potential causal genes for LUAD survival, including aryl hydrocarbon receptor nuclear translocator like 2, semaphorin 3G, serum deprivation-response protein, chloride intracellular channel protein 5, LIM zinc finger domain containing 2, epithelial membrane protein 2, carbonic anhydrase 7 and LOC116437. The results also identified that phosphatidylinositol-3,4,5-trisphosphate-dependent Rac exchange factor 2 may be a potential causal gene for LUSC. Therefore, the results of the present study suggested that there was molecular heterogeneity between these two lung cancer subtypes. Such analysis framework can be extended to other cancer genomics research.

Leveraging Methylation Alterations to Discover Potential Causal Genes Associated With the Survival Risk of Cervical Cancer in TCGA Through a Two-Stage Inference Approach

“…After quality control, we reserved 485,577 DNA methylation CpG sites, three clinical covariates (i.e., age of onset, clinical stage, and tumor status) and up to 190 cervical cancer patients of European ancestry. To avoid numerical instability, we focused on protein-coding genes which had at least 10 methylations within the promoter region and the total gene body ( Liu et al, 2020 ). We also first performed the LMM analysis ( Visscher et al, 2008 ; Yang et al, 2011 ; Zhou et al, 2013 ) for each protein-coding gene based on its methylations and selected genes with the phenotypic variance explained by methylations larger than 1% (corresponding to a correlation coefficient of 10%).…”

“…Second, in our analysis we apply a group of local methylations serving as instrumental variables, which has the potential of higher power because of more variation of expression explained compared to the strategy of applying only a few significantly gene-associated ones ( Zeng and Zhou, 2017 ; Zeng et al, 2021 ). In addition, utilizing local methylation CpG sites for a given gene is also widely seen in gene-based statistical genomics analysis when involving methylations ( Kingsley et al, 2016 ; Loucks et al, 2016 ; Chu and Huang, 2017 ; Huang, 2019 ; Liu et al, 2020 ). However, it has been widely warned in MR studies that incorporating more instrumental variables (e.g., methylations) may have higher risk in violating the third MR assumption (the exclusion restriction assumption; Figure 3 ) due to unknown biological pathways ( Zeng and Zhou, 2019a , b ; Zeng et al, 2019 ; Yuan et al, 2020 ; Liu et al, 2021 ).…”

How can childhood maltreatment affect post-traumatic stress disorder in adult: Results from a composite null hypothesis perspective of mediation analysis

“…More specifically, we would carry out a gene-centric mediation analysis (34,35,40), where each mediation effect test includes one exposure X, multiple methylation mediators M which are per se likely highdimensional, and one outcome Y that may be continuous or binary for n individuals. According to the gene annotation mapping file we define the set of DNA methylation CpG sites for each gene as those located within the entire gene body and 500 bps upstream of the transcription start site (TSS) so that the promoter can be included (61,62). A total of 16,295 genes were analyzed, with the median of the number of DNA methylation CpG sites per gene equal to 16. In the mediation analysis the influence of the exposure X on the outcome Y stands for the direct effect (denoted by c in Figure 1), and the influence of X on M and subsequently M on Y for the indirect effect which is also called the mediation effect (denoted by α β in Figure 1).…”

Significance of screening sensitive methylation sites using whole-genome sequencing in early diagnosis of non-small cell lung cancer

“…Compared with somatic mutation spectrum analysis of early patients, methylation spectrum analysis can achieve higher sensitivity and specificity because: 1) a larger number of markers can be used at the same time to improve sensitivity (15). 2) There are multiple CpG loci in each selected site for the joint query of the region to obtain a "cancer-specific methylation pattern" to improve specificity (16). In addition, methylation profile analysis can be used to distinguish origin tissues from cancer subtypes.…”