Question

Personalised therapy and chemotherapy resistance.

1. Describe some of the steps required to identify PanCancer Driver genes.
2. What are some of the benefits of precision cancer medicine compared to chemotherapy?
3. Describe how acquired resistance to targeted therapies may occur.
4. What are some of the challenges facing precision cancer medicine?

Answer 1

Personalised therapy and chemotherapy resistance.

1. Describe some of the steps required to identify PanCancer Driver genes.
2. What are some of the benefits of precision cancer medicine compared to chemotherapy?
3. Describe how acquired resistance to targeted therapies may occur.
4. What are some of the challenges facing precision cancer medicine?

1. Describe some of the steps required to identify PanCancer Driver genes.

Recent large-scale cancer sequencing studies have discovered many novel cancer driver genes (CDGs) in human cancers. Some studies also suggest that CDG mutations contribute to cancer-associated epigenomic and transcriptomic alterations across many cancer types. Aim to improve our understanding of the connections between CDG mutations and altered cancer cell epigenomes and transcriptomes on PanCancer level and how these connections contribute to the known association between epigenome and transcriptome.

Using multi-omics data including somatic mutation, DNA methylation, and gene expression data of 20 cancer types from The Cancer Genome Atlas (TCGA) project, to conduct a pan-cancer analysis to identify CDGs, when mutated, have strong associations with genome-wide methylation or expression changes across cancer types, which is refered as methylation driver genes (MDGs) or expression driver genes (EDGs), respectively.

Somatic mutation data, DNA methylation 450K array data, and gene-level RNA-seq data of 20 tumor types with at least 100 samples available in all three data types from TCGA. For DNA methylation 450K array data, conducted standard quality control steps removing CpG sites that overlap with known single nuclear polymorphisms (SNPs), sites on the sex chromosomes and sites with missing values for more than 5% of the tumor samples within a tumor type. After these steps, 370,877 CpG sites remained. Further corrected for the type I/II probe bias using the BMIQ algorithm.  

Selection of candidate CDGs across various tumor types

To conduct pan-cancer analysis associating mutation and methylation/expression, within a tumor type, selected CDGs that have mutations in at least 5 samples with matched methylation data or expression data in order to have not-too-sparse numbers in the mutated group. For matched mutation and methylation data, 445 CDGs were selected across the 20 tumor types. Analyzed somatic mutations at the gene level and a gene is considered mutated in a tumor sample as long as there is any mutation in this gene. Within these driver genes, the number of tumor types in which a driver gene was mutated in at least five samples varies from 1 to 15

PanCancer associated CDGs and DNA methylation can be arrived using the following steps:  

CDG be i
Ai be the set of Cancer types genes present in CDG i
Methylation levels range 0 to 1

Further to determine the hyper- or hypo-methylation status per CpG site by the mutation status of CDG iusing the nonparametric Wilcoxon test.

With the Wilcoxon test, genome-wide hyper-methylated sites Si,k+ whose methylation levels are significantly increased at significance level 0.01 in the mutated group comparing to the non-mutated group of CDG i in PanCancer type k.

Hypo-methylated sites be Si,k−

To determine if mutation status of CDG i is significantly associated with genome-wide methylation changes in PanCancer type k, to calculate the p-value pi,k, which is the probability of observing the number of differentially (hyper- or hypo-) methylated sites or more that are associated with the mutation status of CDG i in cancer type k under the null hypothesis that the mutation status of CDG i is not associated with genome-wide methylation changes.  

The 7,019 mutation genes have a similar mutation rate (average number of mutations in a cancer type) with that of the 445 CDGs. The average mutation rate of these 7,019 mutation genes is 0.082 with standard deviation (SD) 0.10 while the average mutation rate of the 445 CDGs is 0.085 with SD = 0.13 (p-value=0.54 from a t-test) were derived using methlation null pool across 20 tumors.

To calculate within PanCancer type, , the number of differentially methylated sites that are associated with the mutation status of the methylationnullgenejnull= 1,…,7019, which form the “methylation null pool”. The p-value pi,k, is then calculated as the proportion of numbers in the “methylation null pool” that is greater than or equal to the observed number of differentially methylated sites that is, , where I(.) is the indicator function.

To investigate the potential selection bias in the “methylation null pool”, we also generated the null distribution of number of genome-wide differentially methylated sites by randomly splitting tumor samples of a tumor type into mutation and non-mutation groups, varying the percentage of mutation from 5 to 40% based on the mutation rate of the TCGA 20 tumor types and calculated numbers of differentially methylated sites between the two groups. We repeated this 10 times for each percentage from 5 to 40%, increasing by 1%. Therefore, we ended up with 360*20 values for the number of differentially methylated sites across 20 tumor types. We found that these numbers are on average much smaller than those from the “methylation null pool” generated using passenger mutations, making the p-values of CDGs more significant. This indicates that there is some association between passenger mutations and global methylation changes that random sampling cannot capture. Therefore, the methylation null pool generated by using the passenger mutations rather than randomly splitting may represent a better null distribution. The MDGs identified this way are those associated with methylation changes beyond what is expected for passenger mutations.

Page Break

CDG i on genome-wide methylation in PanCancer type k is given as:

Enrichment of target gene patterns being hyper-methylated and down-regulated or hypo-methylated and up-regulated, by the 26 overlapping genes. These findings highlight that the dysregulation of chromatin regulation is an important mechanism that amplifies the impact of mutations in CDGs by global methylation and gene expression changes.  

PanCancer analysis examining connections between somatic mutation and DNA methylation/gene expression identified CDGs (32 MDGs and 29 EDGs) whose somatic mutations are associated with genome-wide methylation/expression changes across multiple cancer types. Many of the identified MDGs are either chromatin regulators or the ones that regulate the expression of or physically interact with chromatin regulators. Twenty-six out of the 29 EDGs overlap with the 32 MDGs.