Homework 2

First modify the setup chunk to activate the packages you will need for this HW. Ensure you have tidyverse, knitr and rmarkdown installed. Add statements like library(tidyverse)')

```{r setup}
library(pacman)
p_load(tidyverse) # library(tidyverse)
p_load(knitr)
p_load(rmarkdown)
p_load(janitor)

# add more p_load statements 
# if you need to. Or you can do it in one go with 
# 
# p_load(tidyverse, knitr, rmarkdown, janitor)
```

Instructions

You are expected to edit this R Markdown document, especially the R code chunks. Once you’re done, please change eval=F to eval=T in each chunk in order to make the code run. Most of the R chunks are fill-in-the-blanks, with a few left empty for you to completely fill in.

Your submission for this week will comprise 2 files:

1. This file, after editing to complete the assignments
2. The HTML file obtained by knitting this file

Exercises

1. Fix the code in the following code to ensure they work without errors.

my_variable <-  1:5
my_variable # Change 1 to i)

## [1] 1 2 3 4 5

mean(airquality$Wind) # Wind should be capitalized

table(iris$Sepal.Length) # The actual variable uses ., not _

2. Download the data “clinical_data_breast_cancer_modified.csv” from the data link on Canvas and save it in the data folder of your RStudio Project for this class.

1. Import the data into r calling it breast_cancer.

# Code to import data file here
library(pacman)
p_load(here)
breast_cancer <- rio::import(here("data/clinical_data_breast_cancer_modified.csv"), check.names=TRUE)

## I've added check.names=TRUE to help with removing spaces in the names. 
## 
## Other alternatives are below

# breast_cancer <- data.table::fread("data/clinical_data_breast_cancer_modified.csv")
# breast_cancer <- readr::read_csv('data/clinical_data_breast_cancer_modified.csv')
# breast_cancer <- read.csv('data/clinica_data_breast_cancer_modified.csv', 
#                           stringsAsFactors = FALSE)

2. Check the data types for each of the variables

# Check for data types

glimpse(breast_cancer)

# str(breast_cancer)

3. Convert the Gender ER Status, PR Status and HER2 status variables to factors

# Convert here

# breast_cancer$Gender <- as.factor(breast_cancer$Gender)
# breast_cancer$ER.Status <- as.factor(breast_cancer$ER.Status)

## You can repeat this for the other variables, making sure you spell them properly. Or you can comment 
## proceed as follows:

# breast_cancer <- breast_cancer %>%
#     mutate(Gender = as.factor(Gender), 
#            ER.Status= as.factor(ER.Status),
#            ..... # fill this in with the other variables
#           )

## Comment out one of the two strategies for your submission


## Instead of the repetition of code in the above two strategies, 
## you can use the `across` function to make things more succint

breast_cancer <- breast_cancer %>% 
  mutate(across(c(Gender, ER.Status, PR.Status, HER2.Final.Status), as.factor))

# Not correct (Why?): 
# breast_cancer <- breast_cancer %>%
#   mutate(across(c(Gender,  ends_with("Status")), as.factor))

4. Identify any other variables that might be the wrong type and convert them

# Convert any new variables here. I don't think there is an entirely right
# answer here, but as long as the process was followed, I'm fine with different
# choices

breast_cancer <- breast_cancer %>% 
  mutate(Age.at.Initial.Pathologic.Diagnosis = as.numeric(Age.at.Initial.Pathologic.Diagnosis)) %>% 
  mutate(across(c(Tumor, Node, Metastasis),as.factor))

5. Verify that all the variables in breast_cancer match the modifications you made

# Don't edit this chunk. It is for verification.
breast_cancer %>% 
  dplyr::summarize_all(class) %>% 
  tidyr::pivot_longer(names_to = "Variable", values_to = "Class", cols = everything()) %>% 
  knitr::kable()

Variable	Class
Complete.TCGA.ID	character
Gender	character
Age.at.Initial.Pathologic.Diagnosis	numeric
ER.Status	character
PR.Status	character
HER2.Final.Status	character
Tumor	factor
Tumor..T1.Coded	character
Node	factor
Node.Coded	character
Metastasis	factor
Metastasis.Coded	character
AJCC.Stage	character
Converted.Stage	character
Survival.Data.Form	character
Vital.Status	character
Days.to.Date.of.Last.Contact	integer
Days.to.date.of.Death	integer
OS.event	integer
OS.Time	integer

3. The dataset “clinical_data_breast_cancer_hw.csv” provides clinical and pathological data for 105 breast cancer patients.

1. Read this data into r calling it brca_data.

brca_data <- rio::import(here('data/clinical_data_breast_cancer_hw.csv'))

2. This data is in tidy format, but needs cleaning.
   1. Evaluate whether all the variables are of the correct type. If not, evaluate why, and correct the problem.

glimpse(brca_data)

## Rows: 105
## Columns: 13
## $ Complete.TCGA.ID                    <chr> "TCGA-A2-A0T2", "TCGA-A2-A0CM", "T…
## $ Gender                              <chr> "FEMALE", "FEMALE", "FEMALE", "FEM…
## $ Age.at.Initial.Pathologic.Diagnosis <chr> "66", "40", "48", "56", "38", "57"…
## $ ER.Status                           <chr> "Negative", "Negative", "Negative"…
## $ PR.Status                           <chr> "Negative", "Negative", "Negative"…
## $ HER2.Final.Status                   <chr> "Negative", "Negative", "Negative"…
## $ Tumor                               <chr> "T3", "T2", "T2", "T2", "T3", "T2"…
## $ Node                                <chr> "N3", "N0", "N1", "N1", "N3", "N0"…
## $ Metastasis                          <chr> "M1", "M0", "M0", "M0", "M0", "M0"…
## $ AJCC.Stage                          <chr> "Stage IV", "Stage IIA", "Stage II…
## $ Vital.Status                        <chr> "DECEASED", "DECEASED", "DECEASED"…
## $ Days.to.Date.of.Last.Contact        <int> 240, 754, 1555, 1692, 133, 309, 42…
## $ Days.to.date.of.Death               <int> 240, 754, 1555, 1692, NA, NA, NA, …

## Add code here to correct any problems in the data set

brca_data <- brca_data %>% 
  mutate(across(c(Gender, ends_with("Status"), Tumor:AJCC.Stage), as.factor)) %>% 
  mutate(Age.at.Initial.Pathologic.Diagnosis = as.numeric(Age.at.Initial.Pathologic.Diagnosis))

3. For the molecular results for ER, PR and HER2, any entry that is not Positive or Negative should be changed to a missing value.

This is a rather comprehensive answer, using information from StackOverflow.

This is a non-trivial question since the obvious way doesn’t quite work

brca_data1 <- brca_data %>% 
  mutate(ER.Status = ifelse(ER.Status=='Indeterminate', NA, ER.Status)) %>% 
  mutate(HER2.Final.Status = ifelse(HER2.Final.Status=='Equivocal', NA, HER2.Final.Status))

str(brca_data1 %>% select(ER.Status, HER2.Final.Status))

## 'data.frame':    105 obs. of  2 variables:
##  $ ER.Status        : int  2 2 2 2 2 2 2 2 2 2 ...
##  $ HER2.Final.Status: int  2 2 2 2 2 2 2 2 2 2 ...

OhOh!! The type of the data changed to integers!!

There are several ways to achieve this in a type-safe way. The one I like most is using forcats::fct_recode

brca_data1 <- brca_data %>% 
  mutate(ER.Status = fct_recode(ER.Status, NULL = 'Indeterminate')) %>% 
  mutate(HER2.Final.Status = fct_recode(HER2.Final.Status, NULL = 'Equivocal'))

str(brca_data1 %>% select(ER.Status, HER2.Final.Status))

## 'data.frame':    105 obs. of  2 variables:
##  $ ER.Status        : Factor w/ 2 levels "Negative","Positive": 1 1 1 1 1 1 1 1 1 1 ...
##  $ HER2.Final.Status: Factor w/ 2 levels "Negative","Positive": 1 1 1 1 1 1 1 1 1 1 ...

Note that the forcats package is already loaded into R as part of the tidyverse

Other options include:

brca_data1 <- brca_data %>%
    mutate(ER.Status = recode_factor(ER.Status, Indeterminate = NA_character_)) %>% 
    mutate(HER2.Final.Status = recode_factor(HER2.Final.Status,  
                                             'Equivocal' = NA_character_))

brca_data1 <- brca_data %>%
    mutate(ER.Status = na_if(ER.Status, 'Indeterminate')) %>% 
    mutate(HER2.Final.Status = na_if(HER2.Final.Status, 'Equivocal'))

brca_data1 <- brca_data %>%
    mutate(ER.Status = factor(ER.Status, exclude = 'Indeterminate')) %>% 
    mutate(HER2.Final.Status = factor(HER2.Final.Status, exclude = 'Equivocal'))

The other approach is to create a function that does the transformation and run it on all the columns:

clean_markers <- function(x){
  x <- recode_factor(x, Positive = 'Positive', Negative = 'Negative', .default = NA_character_)
  return(x)
 }

 brca_data2 <- brca_data %>%
mutate(across(c(ER.Status, PR.Status, HER2.Final.Status), clean_markers))

You can verify that both data sets are identical

all(brca_data1 == brca_data2, na.rm=TRUE)

## [1] TRUE

3. Create a new dataset that contains patient ID, ER, PR and HER2 status.

brca_data3 <- select(brca_data, Complete.TCGA.ID, ER.Status, PR.Status, HER2.Final.Status)

# brca_data3 <- brca_data %>% 
#   select(Complete.TCGA.ID, ER.Status, PR.Status, HER2.Final.Status)

4. Creating new variables (what dplyr function will you use for all of these?): mutate

5. Create a variable giving the TNM status of each patient. The T, N and M statuses are given separately. I want a single variable encoded as, for example, “T2N0M0”. [Hint: The function paste is your friend]

brca_data <- brca_data %>%
    mutate(tnm_status = paste(Tumor, Node, Metastasis, sep = ""))
brca_data3$tnm_status <- brca_data$tnm_status

brca_data <- unite(brca_data, tnm_status, c('Tumor','Node', 'Metastasis'), 
                   sep='', remove=FALSE)

6. Create a variable that divides the patients into 3 categories based on molecular profiles [Hint: The function case_when might help]: - Luminal (ER positive and/or PR positive) - HER2 (HER2 positive) - Basal-like (Er PR and HER2 negative)

brca_data3 <- brca_data3 %>%
    mutate(mol_cat = case_when( # fill in the next 4 lines
      (ER.Status =='Positive') | (PR.Status=="Positive") ~ "Luminal",
      (HER2.Final.Status == 'Positive') ~ "HER2",
      (ER.Status=='Negative') & (PR.Status=='Negative') & (HER2.Final.Status=="Negative") ~ 'Basal_like',
      TRUE ~ 'Other'
  ))

7. Create a variable that equals the time of death if they are dead (see the variable Vital.Status) or the time of last contact if they are alive. This is a common computation for survival analysis studies, and is called the overall survival time. [Hint: the function ifelse might be useful]

## Fill this in yourself. Time to start leaving the nest
brca_data <- brca_data %>% 
  mutate(event_time = ifelse(Vital.Status == 'DECEASED', Days.to.date.of.Death, Days.to.Date.of.Last.Contact))
brca_data3$event_time <- brca_data$event_time

Save the cleaned breast cancer dataset as brca_cleaned. You can save this to your computer using saveRDS(brca_cleaned, file="<a filename of your choice>.rds"). We’ll be using this dataset again when we do plots and modeling.

brca_cleaned <- brca_data1
saveRDS(brca_cleaned, file = here('data/brca_data.rds'))