README.md

In Lab-Jaiswal/tidyUkBioBank: What the Package Does (One Line, Title Case)

tidyUkBioBank

tidyUkBioBank is an R-based package that you to easily extract, subset, and parse diagnositic (icd, cancer related, self reported dxs, etc.) from the UK Biobank dataset.

To avoid repetitive calls for standard diagnostic and genetic variables, we created functions to output meaningful columns, such as:

• History of a diagnosis
• Age at which diagnoses were made
• Dates at which diagnoses were made

This package also provides functions you can use to gain a broad understanding of the prevalence of disorders of interest, such as:

• The total number of participants with a diagnosis
• The proportion of males and females afflicted with a certain diagnosis
• The median and mean age patients were first diagnosed

Finally, this package allows you to define a diagnosis based on multiple sources of your choosing, such as:

• icd10 code(s)
• fields : diagnoses_icd10_f41270 _0_0 - 0_222, diagnoses_main_icd10_f41202_0_0 - 0_74, diagnoses_secondary_icd10_f41204_0_0 - 0_183, type_of_cancer_icd10_f40006_0_0 - 0_16, date_of_first_inpatient_diagnosis_icd10_f41280_0_0 - 0_222, date_of_cancer_diagnosis_f40005_0_0 - 0_16
• icd9 code(s)
• fields: diagnoses_icd9_f41271_0_0 - 0_46, diagnoses_secondary_icd9_f41205_0_0 - 0_29, diagnoses_main_icd9_f41203_0_0 - 0_27, type_of_cancer_icd9_f40013_0_0 - 0_14, date_of_first_inpatient_diagnosis_icd9_f41281_0_0 - 0_46, date_of_cancer_diagnosis_f40005_0_0 - 0_16
• cause of death reports
• description_cause_of_death_f40010_0_0, contributory_secondary_ca_f40002_0_0 - 1_14, underlying_primary_cause_of_death_f40001_0_0, date_of_death_f40000_0_0 - 1_0
• self reports
• cancer_code_selfreported_f20001_0_0 - 3_5, noncancer_illness_code_selfreported_f20002_0_0 - 3_33, interpolated_age_of_participant_when_cancer_fist_diagnosed_f2007_0_0 - 3_33, interpolated_year_when_cancer_first_diagnosed_f20006_0_0 - 3_5, interpolated_age_of_participant_when_noncancer_illness_first_diagnosed_f20009_0_0 - 3_33, interpolated_year_when_noncancer_illness_first_diagnosed_f20008_0_0 - 3_5
• date of birth
• fields: year_of_birth_f34_0_0, month_of_birth_f52_0_0
• genetic sex
• genetic_sex_f22001_0_0

Installation and Data Download

Vim

1. Start an srun session
2. Srun -c 16 –mem 128G -t 8:00:00 -A default -p interactive –pty zsh
3. Install and call package
   • devtools::install_github("Lab-Jasiwal/tidyUkBioBank", dependencies = TRUE)
   • library(tidyukbiobank)
4. Call needed dependencies
   • library(tidyverse)
   • library(lubridate)
   • library(magrittr)
5. Download data
   • ukb <- /path/to/data/ukb_phenotype_data_only_relevant_to_tidyukbiobank_Jul_20.rds

ondemand R Studio

1. Go to R Studio Server
   • https://ondemand.scg.stanford.edu/pun/sys/dashboard/batch_connect/sys/rstudio/session_contexts/new
2. Select options:
   • R Version: 4.2.2
   • Number of cores: 16
   • Amount of memory: 258
3. Click "Launch"
4. Call needed dependencies
   • library(tidyukbiobank)
   • library(tidyverse)
   • library(lubridate)
   • library(magrittr)
5. Download data
   • ukb <- /path/to/data/ukb_phenotype_data_only_relevant_to_tidyukbiobank_Jul_20.rds

Prerequisites

Obtaining Phenotypic Data from the UKBioBank

1. Make a UKB fileset a. Download and decrypt your data from the Uk BioBank b. Run the following commands to create a fileset:

    ukb_unpack ukbxxxx.enc key
    ukb_conv ukbxxxx.enc_ukb r
    ukb_conv ukbxxxx.enc_ukb docs
   

   1. Make a UKB dataset a. Install the ukbtools package:

      devtools::install_github("kenhanscombe/ukbtools", dependencies = TRUE)

b. Run ukbtools ukb_df command, which requires the stem of your fileset and its path:

    ukb_data <- ukb_df("ukbxxxx", path = "/full/path/to/my/data")

The ukb_df() function returns a dataframe with usable column names. This command may take several minutes and, depending on the amount of memory available, you may need to split the ukbxxxx.tab dataframe into several peices. I used the following commands:

In bash:

    cat ukbxxxxx.tab | parallel -j 16 --header : --pipe -N10000 'cat >subset_{#}.tab' 
    for x in ./*.tab; do
      mkdir "${x%.*}" && mv "$x" "${x%.*}"
    done
    for i in ./* # iterate over all files in current dir
      do
       if [ -d "$i" ] # if it's a directory
        then
          cp ukbxxxx.r "$i" # copy ukbxxxx.r into it
          cp ukbxxxx.log "$i" # copy ukbxxxx.log into it
          cp ukbxxxx.html "$i" # copy ukbxxxx.html into it
       fi
     done
     dirname $(ls */*.tab) | xargs -I % bash -c "mv %/%.tab %/ukbxxxx.tab"

In R:

    library(ukbtools)
    library(magrittr)
    library(tidyverse) 
    subset_dir <- "file/path/with/subsetted/data"
    subset_list <- list.files(subset_dir, pattern="subset_*") %>% sort
    subset_list_path <- str_c(subset_dir, subset_list, sep="/")

    read_df <- function(subsets_dir) {
           ukb_object <- ukb_df("ukbxxxx", path=subsets_dir)
           tibble(ukb_object) 
    }

    ukb_full <- map(subset_list_path, read_df) %>% bind_rows()
    write_rds(ukb_full, "ukb_phen_data_final_todays_date.rds")

§ Full details on creating a UKB fileset: https://biobank.ctsu.ox.ac.uk/~bbdatan/Accessing_UKB_data_v2.3.pdf § Full details on making a UKB dataset: https://github.com/kenhanscombe/ukbtools § Full details on downloading UKB genetic data: https://biobank.ndph.ox.ac.uk/ukb/ukb/docs/ukbgene_instruct.html § Full details on parallel_fetch.sh script: https://github.com/neurodatascience/ukbm § Additional thread on splitting the tab data: https://github.com/kenhanscombe/ukbtools/issues/1

Functions

There are 29 functions in tidyukbiobank. Luckily, you only need to worry about 5 of them: parse_get_SR_table_input diagnoses_table diagnoses_counts diagnoses_ages * diagnoses_dates

parse_get_SR_table

Arguments

disease: string either from coding6/ coding3 OR string with a description of the disease of interest

cancer: whether the diagnoses is a cancer diagnosis TRUE or FALSE For cancer diagnoses, put just the body part and then TRUE. For example: parse_get_SR_table_input("breast", TRUE)

Example:

diagnoses_table(), diagnoses_counts(), diagnoses_ages(), diagnoses_dates()

Arguments

Mandatory: * dataframe: name of the dataframe with the ukbb data

Optional: icd_code_list: a list of icds the patient was dxd with during an inpt hospital stay cause_of_death: a list of icds associated with the patient’s primary or secondary cause of death * self_reported: a code or string corresponding to either coding6 or codign3

Example

diagnoses_table

Output: A table containing eids and columns filled with 1’s and 0’s for if the eid has a record of the icd, cod, or self diagnosed code.

Example:

diagnoses_counts

Outputs: A counts table containing the count (female, male, and combined), % female, % male, mean and median age dxd (for females, males, and combined) for all codes provided

Example:

Notice the following: 1. The numbers associated with ICD codes L93-7100 do not add up to the numbers in Combined_ICD_Codes. This is to be expected since there are individuals who have obtained multiple ICD codes. 2. In 6854, 7100, M321, M328, and M329 the Total column contains more people than the male and female columns combined. This is because there are approximately 14,000 people in the UK biobank with NA for genetic_sex_f22001_0_0

diagnoses_dates

Output: A table containing the eids and date dxd for each icd/ cause of death/ self reported code given. If any are eids missing that show a 1 in using diagnoses_table, it is because no date of diagnoses was provided.

Example:

diagnoses_ages

Output: A table containing the eids and age dxd for each icd/ cause of death/ self reported code given. If any are eids missing that show a 1 in using diagnoses_table, it is because no date of diagnoses was provided.

Example:

Citations

Hanscombe KB, Coleman J, Traylor M, Lewis CM (e-print 158113). “ukbtools: An R package to manage and query UK Biobank data.” bioRxiv. .

Lab-Jaiswal/tidyUkBioBank documentation built on Sept. 10, 2023, 5:13 p.m.