cancer gene census copy number

Ming Tang

January 2, 2018

Set up

knitr will force changing current working directory: https://philmikejones.wordpress.com/2015/05/20/set-root-directory-knitr/
https://github.com/yihui/knitr/issues/277

or use the ezkintr package https://deanattali.com/blog/ezknitr-package/

library(knitr)
library(here)

## here() starts at /Users/mtang1/projects/mixing_histology_lung_cancer

root.dir<- here()
opts_knit$set(root.dir = root.dir)

read in the data

I am going to check the COSMIC database on cancer genes. Specifically, I want to know which cancer-related genes are found amplified and which are deleted.

The data can be downloaded from http://cancer.sanger.ac.uk/cosmic/census?genome=37#cl_search

library(tidyverse)

## Loading tidyverse: ggplot2
## Loading tidyverse: tibble
## Loading tidyverse: tidyr
## Loading tidyverse: readr
## Loading tidyverse: purrr
## Loading tidyverse: dplyr

## Conflicts with tidy packages ----------------------------------------------

## filter(): dplyr, stats
## lag():    dplyr, stats

library(janitor)
library(stringr)
cancer_gene_census<- read_csv("data/COSMIC_Cancer_gene_Census/cancer_gene_census.csv", col_names = T)

## Parsed with column specification:
## cols(
##   .default = col_character(),
##   `Entrez GeneId` = col_integer(),
##   Tier = col_integer()
## )

## See spec(...) for full column specifications.

tidy the data with janitor::clean_names(), dplyr::unnest

The column names have spaces, and in many columns there are mulitple strings separated by ,.

## dplyr after 0.7.0 use pull to get one column out as a vector, I was using .$
# https://stackoverflow.com/questions/27149306/dplyrselect-one-column-and-output-as-vector

#cancer_gene_census %>% pull(`Gene Symbol`) %>% unique()
#cancer_gene_census %>% distinct(`Gene Symbol`)

## extract genes that are amplified or deleted in cancers. in the `mutation types` column search A for amplificaton and D for large deletion.

## use janitor::clean_names() to clean the column names with space.
cancer_gene_census %>% 
        clean_names() %>%
        mutate(mutation_type =strsplit(as.character(mutation_types), ",")) %>% 
        unnest(mutation_type) %>% tabyl(mutation_type)

##    mutation_type   n      percent valid_percent
## 1              N   1 0.0008257638  0.0008264463
## 2              A   8 0.0066061107  0.0066115702
## 3              D   8 0.0066061107  0.0066115702
## 4              F 139 0.1147811726  0.1148760331
## 5            Mis  68 0.0561519405  0.0561983471
## 6              N 147 0.1213872832  0.1214876033
## 7              O  37 0.0305532618  0.0305785124
## 8              S  76 0.0627580512  0.0628099174
## 9              T  19 0.0156895128  0.0157024793
## 10             A  18 0.0148637490  0.0148760331
## 11             D  35 0.0289017341  0.0289256198
## 12             F  32 0.0264244426  0.0264462810
## 13        F; Mis   1 0.0008257638  0.0008264463
## 14             M   3 0.0024772915  0.0024793388
## 15           Mis 240 0.1981833196  0.1983471074
## 16        Mis. N   1 0.0008257638  0.0008264463
## 17             N  24 0.0198183320  0.0198347107
## 18             O   7 0.0057803468  0.0057851240
## 19  Promoter Mis   1 0.0008257638  0.0008264463
## 20             S   2 0.0016515277  0.0016528926
## 21             T 343 0.2832369942  0.2834710744
## 22          <NA>   1 0.0008257638            NA

It turns out that single mutation_types column is more messy than you think… Multiple entries of A and D in different rows. spaces in the column are the devil.

trim the spaces with stringr::str_trim()

Abbreviations can be found http://cancer.sanger.ac.uk/cosmic/census?genome=37#cl_download

# trim the white space
cancer_gene_census %>% 
        clean_names() %>%
        mutate(mutation_type =strsplit(as.character(mutation_types), ",")) %>% 
        unnest(mutation_type) %>% 
        mutate(mutation_type = str_trim(mutation_type)) %>%
        filter(mutation_type == "A" | mutation_type == "D") %>% tabyl(mutation_type)

##   mutation_type  n   percent
## 1             A 26 0.3768116
## 2             D 43 0.6231884

cancer_gene_census %>% 
        clean_names() %>%
        mutate(mutation_type =strsplit(as.character(mutation_types), ",")) %>% 
        unnest(mutation_type) %>% 
        mutate(mutation_type = str_trim(mutation_type)) %>%
        filter(mutation_type == "A" | mutation_type == "D") %>% count(mutation_type)

## # A tibble: 2 x 2
##   mutation_type     n
##           <chr> <int>
## 1             A    26
## 2             D    43

Sanity check

according to the website http://cancer.sanger.ac.uk/cosmic/census?genome=37#cl_sub_tables there are should be 40 deletions and 24 amplifications while I am getting 43 and 26, respectively.

cancer_gene_census %>% 
        clean_names() %>%
        mutate(mutation_type =strsplit(as.character(mutation_types), ",")) %>% 
        unnest(mutation_type) %>% 
        mutate(mutation_type = str_trim(mutation_type)) %>%
        filter(mutation_type == "A") 

## # A tibble: 26 x 21
##    gene_symbol
##          <chr>
##  1        AKT2
##  2        AKT3
##  3         ALK
##  4       CCNE1
##  5      DROSHA
##  6        EGFR
##  7       ERBB2
##  8         ERG
##  9        FLT4
## 10        GRM3
## # ... with 16 more rows, and 20 more variables: name <chr>,
## #   entrez_geneid <int>, genome_location <chr>, tier <int>,
## #   hallmark <chr>, chr_band <chr>, somatic <chr>, germline <chr>,
## #   tumour_types_somatic <chr>, tumour_types_germline <chr>,
## #   cancer_syndrome <chr>, tissue_type <chr>, molecular_genetics <chr>,
## #   role_in_cancer <chr>, mutation_types <chr>,
## #   translocation_partner <chr>, other_germline_mut <chr>,
## #   other_syndrome <chr>, synonyms <chr>, mutation_type <chr>

I checked by eyes, AKT3 and GRM3 are missing from the table online http://cancer.sanger.ac.uk/cosmic/census/tables?name=amp but when I checked the downloaded table, both AKT3 and GRM3 has Amplification in the mutation types column. I am bit confused, but for now I will stick to the downloaded table.

It teaches me an important lesson.

1. Do not trust data (blindly) from any resource even COSMIC.
2. Data are always messy. Tidying data is the big job.
3. Becareful with the data, check with sanity. Browse the data with your eyes may reveal some unexpected things.

oncogenes are amplified and tumor suppressor genes are always deleted?

In cancer, we tend to think oncogenes are amplified and tumor suppressor genes are deleted to promote tumor progression. Is it true in this setting?

#devtools::install_github("haozhu233/kableExtra")
library(kableExtra)
library(knitr)
cancer_gene_census %>% 
        clean_names() %>%
        mutate(mutation_type =strsplit(as.character(mutation_types), ",")) %>% 
        unnest(mutation_type) %>% 
        mutate(mutation_type = str_trim(mutation_type)) %>%
        filter(mutation_type == "A" | mutation_type == "D") %>% 
        mutate(role = strsplit(as.character(role_in_cancer), ",")) %>%
        unnest(role) %>%
        mutate(role = str_trim(role)) %>%
        filter(role == "TSG" | role == "oncogene") %>%
        count(role, mutation_type) %>% kable()

role	mutation_type	n
oncogene	A	25
oncogene	D	7
TSG	A	3
TSG	D	42

what are those genes?

cancer_gene_census %>% 
        clean_names() %>%
        mutate(mutation_type =strsplit(as.character(mutation_types), ",")) %>% 
        unnest(mutation_type) %>% 
        mutate(mutation_type = str_trim(mutation_type)) %>%
        filter(mutation_type == "A" | mutation_type == "D") %>% 
        mutate(role = strsplit(as.character(role_in_cancer), ",")) %>%
        unnest(role) %>%
        mutate(role = str_trim(role)) %>%
        filter(role == "TSG" | role == "oncogene") %>%
        filter((role == "TSG" & mutation_type == "A") | (role == "oncogene" & mutation_type == "D")) %>%
        select(gene_symbol, role_in_cancer, role, mutation_types, mutation_type) %>%
        kable(format = "html", booktabs = T, caption = "TSG and oncogene") %>%
        kable_styling(latex_options = c("striped", "hold_position"),
                full_width = F)

TSG and oncogene

gene_symbol	role_in_cancer	role	mutation_types	mutation_type
APOBEC3B	oncogene, TSG	oncogene	D	D
BIRC3	oncogene, TSG, fusion	oncogene	D, F, N, T, Mis	D
DROSHA	TSG	TSG	A, Mis, N, F	A
GPC3	oncogene, TSG	oncogene	D, Mis, N, F, S	D
KDM6A	oncogene, TSG	oncogene	D, N, F, S	D
MAP2K4	oncogene, TSG	oncogene	D, Mis, N	D
NKX2-1	oncogene, TSG	TSG	A	A
NTRK1	oncogene, TSG, fusion	TSG	T, A	A
PAX5	oncogene, TSG, fusion	oncogene	T, Mis, D, F, S	D
WT1	oncogene, TSG, fusion	oncogene	D, Mis, N, F, S, T	D

majority of the genes from the output can function as either oncogene or tumor suppressor genes (TSG), which is not totally suprising. Cancer is such a complex disease that a function of the gene is dependent on the context (e.g. cancer types). Interestingly, DROSHA is the only TSG and is amplified.