Last updated: 2025-12-28
Checks: 7 0
Knit directory:
genomics_ancest_disease_dispar/
This reproducible R Markdown analysis was created with workflowr (version 1.7.1). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.
Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.
Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.
The command set.seed(20220216) was run prior to running
the code in the R Markdown file. Setting a seed ensures that any results
that rely on randomness, e.g. subsampling or permutations, are
reproducible.
Great job! Recording the operating system, R version, and package versions is critical for reproducibility.
Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.
Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.
Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.
The results in this page were generated with repository version 410a36a. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.
Note that you need to be careful to ensure that all relevant files for
the analysis have been committed to Git prior to generating the results
(you can use wflow_publish or
wflow_git_commit). workflowr only checks the R Markdown
file, but you know if there are other scripts or data files that it
depends on. Below is the status of the Git repository when the results
were generated:
Ignored files:
Ignored: .DS_Store
Ignored: .Rproj.user/
Ignored: .venv/
Ignored: analysis/.DS_Store
Ignored: ancestry_dispar_env/
Ignored: data/.DS_Store
Ignored: data/cohort/
Ignored: data/gbd/.DS_Store
Ignored: data/gbd/IHME-GBD_2021_DATA-d8cf695e-1.csv
Ignored: data/gbd/IHME-GBD_2023_DATA-73cc01fd-1.csv
Ignored: data/gbd/ihme_gbd_2019_global_disease_burden_rate_all_ages.csv
Ignored: data/gbd/ihme_gbd_2019_global_paf_rate_percent_all_ages.csv
Ignored: data/gbd/ihme_gbd_2021_global_disease_burden_rate_all_ages.csv
Ignored: data/gbd/ihme_gbd_2021_global_paf_rate_percent_all_ages.csv
Ignored: data/gwas_catalog/
Ignored: data/icd/.DS_Store
Ignored: data/icd/IHME_GBD_2019_COD_CAUSE_ICD_CODE_MAP_Y2020M10D15.XLSX
Ignored: data/icd/IHME_GBD_2019_NONFATAL_CAUSE_ICD_CODE_MAP_Y2020M10D15.XLSX
Ignored: data/icd/IHME_GBD_2021_COD_CAUSE_ICD_CODE_MAP_Y2024M05D16.XLSX
Ignored: data/icd/IHME_GBD_2021_NONFATAL_CAUSE_ICD_CODE_MAP_Y2024M05D16.XLSX
Ignored: data/icd/UK_Biobank_master_file.tsv
Ignored: data/icd/cdc_valid_icd10_Sep_23_2025.xlsx
Ignored: data/icd/cdc_valid_icd9_Sep_23_2025.xlsx
Ignored: data/icd/hp_umls_mapping.csv
Ignored: data/icd/lancet_conditions_icd10.xlsx
Ignored: data/icd/manual_disease_icd10_mappings.xlsx
Ignored: data/icd/mondo_umls_mapping.csv
Ignored: data/icd/phecode_international_version_unrolled.csv
Ignored: data/icd/phecode_to_icd10_manual_mapping.xlsx
Ignored: data/icd/semiautomatic_ICD-pheno.txt
Ignored: data/icd/semiautomatic_ICD-pheno_UKB_subset.txt
Ignored: human_dictionary/
Ignored: igsr_populations.tsv
Ignored: output/.DS_Store
Ignored: output/abstracts/
Ignored: output/doccano/
Ignored: output/fulltexts/
Ignored: output/gwas_cat/
Ignored: output/gwas_cohorts/
Ignored: output/icd_map/
Ignored: output/trait_ontology/
Ignored: pubmedbert-cohort-ner-model/
Ignored: pubmedbert-cohort-ner/
Ignored: r-spacyr/
Ignored: renv/
Ignored: venv/
Untracked files:
Untracked: code/extract_cdc_meta.R
Untracked: code/figure_4a.R
Untracked: code/poster_figures.R
Untracked: code/umls_ontology.R
Untracked: data/cdc/
Untracked: data/icd/2025AA/
Untracked: data/icd/umls-2025AA-mrconso.zip
Untracked: figures/
Untracked: visualization.Rdata
Unstaged changes:
Modified: analysis/disease_inves_by_ancest.Rmd
Modified: analysis/get_full_text.Rmd
Modified: analysis/group_cancer_diseases.Rmd
Modified: analysis/gwas_to_gbd.Rmd
Modified: analysis/index.Rmd
Modified: analysis/level_1_disease_group_non_cancer.Rmd
Modified: analysis/level_2_disease_group.Rmd
Modified: analysis/manual_trait_map_icd10.Rmd
Modified: analysis/map_trait_to_icd10.Rmd
Modified: analysis/missing_cohort_info.Rmd
Modified: analysis/replication_ancestry_bias.Rmd
Modified: analysis/trait_ontology_categorization.Rmd
Modified: code/custom_plotting.R
Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.
These are the previous versions of the repository in which changes were
made to the R Markdown
(analysis/text_for_cohort_labels.Rmd) and HTML
(docs/text_for_cohort_labels.html) files. If you’ve
configured a remote Git repository (see ?wflow_git_remote),
click on the hyperlinks in the table below to view the files as they
were in that past version.
| File | Version | Author | Date | Message |
|---|---|---|---|---|
| Rmd | 410a36a | IJbeasley | 2025-12-28 | Include GWAS Catalog cohorts in grep search for cohort sentences |
| html | 238486e | IJbeasley | 2025-10-24 | Build site. |
| Rmd | 0d8b872 | IJbeasley | 2025-10-24 | Cleaning up abstract collecting code again |
| html | 2afc108 | IJbeasley | 2025-10-24 | Build site. |
| Rmd | 748dac2 | IJbeasley | 2025-10-24 | Cleaning up abstract collecting |
library(stringr)
library(readxl)
library(dplyr)
library(stringi)
library(rentrez)
library(xml2)
library(jsonlite)
# Improve sentence recognition
# library(reticulate)
# Path to the Python inside your venv
# python_path <- file.path(here::here(), "venv", "bin", "python") # Mac/Linux
# Sys.setenv(SPACY_PYTHON = python_path)
#
# use_python(Sys.getenv("SPACY_PYTHON",
# unset = "r-spacyr"), required = TRUE)
#
# # Check configuration
# py_config()
#
# py_module_available("spacy")
#
# library(spacyr)
# spacy_initialize(model = "en_core_web_sm")
# library(spacyr)
# reticulate::virtualenv_create("r-spacyr", python = python_exe)
# spacy_install(version = "apple")
# spacy_download_langmodel("en_core_web_sm")
library(tokenizers)## Step 1:
# get only disease studies
gwas_study_info <- data.table::fread(here::here("output/gwas_cat/gwas_study_info_trait_group_l2.csv"))
gwas_study_info = gwas_study_info |>
dplyr::rename_with(~ gsub(" ", "_", .x))
gwas_study_info =
gwas_study_info |>
dplyr::filter(DISEASE_STUDY == T) |>
dplyr::select(-COHORT)gwas_study_info_cohort =
data.table::fread(here::here("output/gwas_cohorts/gwas_cohort_name_corrected.csv"))
gwas_study_info_cohort =
gwas_study_info_cohort |>
dplyr::rename_with(~ gsub(" ", "_", .x))
gwas_study_info_cohort =
gwas_study_info_cohort |>
select(STUDY_ACCESSION,
COHORT) |>
distinct()
gwas_study_info =
left_join(gwas_study_info,
gwas_study_info_cohort,
by = "STUDY_ACCESSION"
)gwas_ancest_info <- data.table::fread(here::here("data/gwas_catalog/gwas-catalog-v1.0.3.1-ancestries-r2025-07-21.tsv"),
sep = "\t",
quote = "")
gwas_ancest_info = gwas_ancest_info |>
dplyr::rename_with(~ gsub(" ", "_", .x))
gwas_ancest_info =
gwas_ancest_info |>
select(STUDY_ACCESSION,
BROAD_ANCESTRAL_CATEGORY,
COUNTRY_OF_RECRUITMENT) |>
distinct() |>
group_by(STUDY_ACCESSION) |>
summarise(
BROAD_ANCESTRAL_CATEGORY = paste(
unique(
unlist(strsplit(BROAD_ANCESTRAL_CATEGORY, split = "\\|"))
),
collapse = "|"
),
COUNTRY_OF_RECRUITMENT = paste(
unique(
unlist(strsplit(COUNTRY_OF_RECRUITMENT, split = "\\|"))
),
collapse = "|"
)
)
gwas_study_info =
left_join(gwas_study_info,
gwas_ancest_info,
by = "STUDY_ACCESSION"
)gwas_study_info <-
gwas_study_info |>
#filter(COHORT != "") |>
select(PUBMED_ID,
COHORT,
DATE,
BROAD_ANCESTRAL_CATEGORY,
COUNTRY_OF_RECRUITMENT) |>
distinct() |>
group_by(PUBMED_ID,
DATE,
BROAD_ANCESTRAL_CATEGORY,
COUNTRY_OF_RECRUITMENT) |>
summarise(
COHORT = paste(
unique(
unlist(strsplit(COHORT, split = "\\|"))
),
collapse = "|"
)
)`summarise()` has grouped output by 'PUBMED_ID', 'DATE',
'BROAD_ANCESTRAL_CATEGORY'. You can override using the `.groups` argument.pmids = gwas_study_info$PUBMED_ID
cohort = gwas_study_info$COHORT
date = gwas_study_info$DATE
country = gwas_study_info$COUNTRY_OF_RECRUITMENT
ancestry = gwas_study_info$BROAD_ANCESTRAL_CATEGORY
# how many papers without cohort information
gwas_study_info |>
filter(COHORT == "") |>
nrow()[1] 4396# papers for which there is cohort information
gwas_study_info |>
filter(COHORT != "") |>
nrow()[1] 1236# Example PMIDs
rows_with_cohort = which(gwas_study_info$COHORT != "")
rows_without_cohort = which(gwas_study_info$COHORT == "")
rows =
c(rows_with_cohort,
sample(rows_without_cohort, length(rows_with_cohort))
)
pmids = pmids[rows]
cohort = cohort[rows]
names(cohort) = pmids
date = date[rows]
names(date) = pmids
country = country[rows]
names(country) = pmids
ancestry = ancestry[rows]
names(ancestry) = pmids# this xlsx was built from looking at acrynyms / cohort names in the gwas catalog
# and finding the corresponding full names / details of cohorts
cohort_names <- readxl::read_xlsx(here::here("data/cohort/cohort_desc.xlsx"),
sheet = 1) |>
mutate(across(everything(),
~stringr::str_replace_all(.x,
pattern = "\u00A0",
replacement = " "))) New names:
• `` -> `...15`cohort_full_names = cohort_names$full_name[!is.na(cohort_names$full_name)]
cohort_full_names <- str_trim(cohort_full_names)
cohort_full_names <- iconv(cohort_full_names, to = "UTF-8")
cohort_full_names <- gsub("[\u00A0\r\n]", " ", cohort_full_names) # replace non-breaking spaces, CR, LF with space
cohort_full_names <- str_squish(cohort_full_names) # trims and removes extra spaces
# sort by length of name (longest first) to match longest names first
cohort_full_names <- cohort_full_names[order(-nchar(cohort_full_names))]
cohort_full_names <- cohort_full_names[cohort_full_names != "Not Reported"]
cohort_abbr_names = cohort_names$cohort[!is.na(cohort_names$cohort)]
cohort_abbr_names <- str_trim(cohort_abbr_names)
cohort_abbr_names <- iconv(cohort_abbr_names, to = "UTF-8")
cohort_abbr_names <- gsub("[\u00A0\r\n]", " ", cohort_abbr_names) # replace non-breaking spaces, CR, LF with space
cohort_abbr_names <- str_squish(cohort_abbr_names) # trims and removes extra spaces
# remove abbreviations that are too short
cohort_abbr_names <- cohort_abbr_names[nchar(cohort_abbr_names) >= 4]
small_abbr_to_keep <- c("C4D",
"BBJ",
"UKB",
"MVP",
"TWB",
"QBB",
"MEC"
)
cohort_abbr_names <- unique(c(cohort_abbr_names,
small_abbr_to_keep
))
cohort_abbr_names <- cohort_abbr_names[!str_detect(pattern = "\\?", cohort_abbr_names)]
# sort by length of name (longest first) to match longest names first
cohort_abbr_names <- cohort_abbr_names[order(-nchar(cohort_abbr_names))]
# add cohort names from GWAS catalog not yet added to data-dictionary
gwas_cat_cohorts = gwas_study_info_cohort$COHORT
gwas_cat_cohorts = unlist(strsplit(gwas_cat_cohorts, "\\|"))
gwas_cat_cohorts = gwas_cat_cohorts[!(gwas_cat_cohorts %in% c("", "other", "multiple"))]
cohort_abbr_names = unique(cohort_abbr_names,
gwas_cat_cohorts)set_entrez_key(Sys.getenv('NCBI_API_KEY'))
get_pubmed_abstracts <- function(pmids,
batch_size = 200,
verbose = TRUE) {
n <- length(pmids)
abstracts <- setNames(rep("MISSING",
n),
pmids
) # initialize result
# Split PMIDs into batches
batches <- split(pmids,
ceiling(seq_along(pmids)/batch_size)
)
for(i in seq_along(batches)) {
batch_pmids <- batches[[i]]
if(verbose) message(sprintf("Fetching batch %d of %d (%d PMIDs)...",
i,
length(batches),
length(batch_pmids)
)
)
# Fetch XML
xml_data <- entrez_fetch(db = "pubmed",
id = paste(batch_pmids,
collapse = ","),
rettype = "xml",
parsed = FALSE
)
doc <- read_xml(xml_data)
articles <- xml_find_all(doc, ".//PubmedArticle")
for(article in articles) {
pmid_node <- xml_find_first(article,
".//PMID")
pmid <- xml_text(pmid_node)
abstract_nodes <- xml_find_all(article,
".//AbstractText")
if(length(abstract_nodes) > 0) {
abstracts[pmid] <- paste(xml_text(abstract_nodes),
collapse = " ")
}
}
}
return(abstracts)
}
pmids = unique(pmids)
abstracts <- get_pubmed_abstracts(pmids)
pmids <- pmids[abstracts != "MISSING"]
date <- date[abstracts != "MISSING"]
cohort <- cohort[abstracts != "MISSING"]
abstracts <- abstracts[abstracts != "MISSING"]
# Loop through abstracts and write each to a file
for (i in seq_along(abstracts)) {
file_name <- paste0(here::here("output/abstracts/"), pmids[i], ".txt")
writeLines(abstracts[i], file_name)
}abstract_files <- list.files(here::here("output/abstracts/"),
pattern = "*.txt",
full.names = FALSE
)
pmids_with_abstracts = gsub("\\.txt$", "", abstract_files)
all_pmids = gwas_study_info$PUBMED_ID |>
unique()
missing_abstracts = setdiff(all_pmids,
pmids_with_abstracts)
print("Number of missing abstracts:")[1] "Number of missing abstracts:"length(missing_abstracts)[1] 42# library(openalexR)
#
# # get information on these papers from openAlex
# oa_example <-
# oa_fetch(entity = "works",
# pmid = missing_abstracts,
# abstract = TRUE)
#
# oa_example |>
# select(doi, abstract) |>
# distinct()abstracts <- sapply(abstract_files,
function(file) {
readLines(here::here(paste0("output/abstracts/",file)),
warn = FALSE) |>
paste(collapse = " ")
}
)
pmids = pmids_with_abstracts
cohort = cohort[pmids]
date = date[pmids]
country = country[pmids]extract_cohort_sentences <- function(text_vector,
cohort_names,
ignore_case) {
results <- lapply(seq_along(text_vector), function(i) {
abstract <- text_vector[i]
# Split abstract into sentences
sentences <- unlist(tokenize_sentences(abstract))
# For each sentence, find all matching cohort names
lapply(seq_along(sentences), function(s) {
sentence <- sentences[s]
# Identify cohort names present in this sentence
matched_cohorts <- cohort_names[str_detect(sentence,
regex(cohort_names,
ignore_case = ignore_case))]
data.frame(
abstract_id = i,
sentence_id = s,
sentence = sentence,
has_cohort = length(matched_cohorts) > 0,
COHORT = if (length(matched_cohorts) > 0) str_flatten(unique(matched_cohorts), collapse = "|", na.rm = T) else "",
stringsAsFactors = FALSE
)
}) |> bind_rows()
})
bind_rows(results)
}
cohort_sentences_df <- extract_cohort_sentences(abstracts,
cohort_full_names,
ignore_case = TRUE
)
cohort_sentences_df_p2 = extract_cohort_sentences(abstracts,
cohort_abbr_names,
ignore_case = FALSE
)
cohort_sentences_df =
bind_rows(cohort_sentences_df,
cohort_sentences_df_p2
)
cohort_sentences_df =
cohort_sentences_df |>
distinct()# set the ratio of sentences with/without cohort names
# ratio = 1
#
# # number of sentences with cohort names
# n_tp_sentences = nrow(cohort_sentences_df |>
# filter(COHORT != "")
# )
# cohort_sentences_df =
# cohort_sentences_df |>
# mutate(has_cohort = ifelse(COHORT == "", FALSE, TRUE)) |>
# group_by(has_cohort) |>
# slice_sample(n = ratio*n_tp_sentences,
# replace = FALSE
# )
filtered_cohort_sentences_df =
cohort_sentences_df |>
filter(COHORT != "")
control_cohort_sentences_df <-
cohort_sentences_df |>
filter(grepl("cohort|consortium|study|population|registry|biobank|corsortia",
sentence,
ignore.case = TRUE)
) |>
slice_sample(n = 1000)
filtered_cohort_sentences_df =
bind_rows(filtered_cohort_sentences_df,
control_cohort_sentences_df
) |>
distinct()
filtered_cohort_sentences_df =
filtered_cohort_sentences_df |>
group_by(abstract_id, sentence_id, sentence) |>
summarise(
COHORT = str_flatten(unique(COHORT), collapse = "|", na.rm = T)
) |>
ungroup()`summarise()` has grouped output by 'abstract_id', 'sentence_id'. You can
override using the `.groups` argument.abstract_ids = filtered_cohort_sentences_df$abstract_id |> unique() |> sort()
pmids = pmids[abstract_ids]
abstracts = abstracts[abstract_ids]
date = date[abstract_ids]
cohort = cohort[abstract_ids]# file path for intermediate json file output
json_file = here::here("output/doccano/abstracts_with_cohort_info.json")
# file path for final jsonl file output
jsonl_file = here::here("output/doccano/abstracts_with_cohort_info.jsonl")
convert_to_doccano_json_sentence_level <- function(pmids,
date,
cohort,
country,
cohort_sentences_df) {
# set up json list
doccano_list <- list()
example_id <- 1
for(current_sentence in cohort_sentences_df$sentence) {
# Filter cohort sentences that match this sentence (safe matching)
df <- cohort_sentences_df |>
dplyr::filter(sentence == current_sentence)
# abstract_id <- df$abstract_id
# matched_cohort <- df$COHORT
for (i in seq_len(nrow(df))) {
matched_cohort <- df$COHORT[i]
abstract_id <- df$abstract_id[i]
#browser()
if(matched_cohort == ""){
doccano_list[[example_id]] <- list(
text = current_sentence,
pubmed_id = pmids[abstract_id],
date = date[abstract_id],
country = country[abstract_id],
gwas_cat_cohort_label = cohort[abstract_id],
label = list()
)
} else {
# Find the location of all matches of the cohort name in the sentence
if(grepl("\\|", matched_cohort)) {
# If multiple cohort names, separate
matched_cohort <- unlist(
strsplit(matched_cohort,
split = "\\|")
)
match_locations <- list()
for(current_matched_cohort in matched_cohort){
matches <- str_locate_all(current_sentence,
fixed(current_matched_cohort,
ignore_case = TRUE)
)[[1]]
match_locations <- append(match_locations,
list(matches)
)
}
# Combine all match locations into a single matrix
matches <- do.call(rbind, match_locations)
} else {
matches <- str_locate_all(current_sentence,
fixed(matched_cohort,
ignore_case = TRUE)
)[[1]]
}
# Convert matches to 0-based indexing (for doccano)
matches[, "start"] <- matches[, "start"] - 1
# Turn match locations into entity list
entities <- list()
for(k in seq_len(nrow(matches))) {
entities <- append(entities, list(list(
start_offset = matches[k, "start"],
end_offset = matches[k, "end"],
label = "COHORT"
)))
}
# Create Doccano JSON entry
doccano_list[[example_id]] <- list(
# id = example_id,
text = current_sentence,
pubmed_id = pmids[abstract_id],
date = date[abstract_id],
country = country[abstract_id],
gwas_cat_cohort_label = cohort[abstract_id],
label = entities
)
}
}
example_id <- example_id + 1
}
# Suppose each element of json_list has 'labels' as a list of named lists
doccano_list <- lapply(doccano_list, function(x) {
x$label <- lapply(x$label, function(l) {
# convert named list to vector/list format [start, end, label]
c(l$start_offset, l$end_offset, l$label)
})
x
})
return(doccano_list)
}
# Create JSON as a list
json_list <- convert_to_doccano_json_sentence_level(pmids = pmids,
date = date,
cohort = cohort,
country = country,
filtered_cohort_sentences_df)
# Write to JSON file
writeLines(toJSON(json_list,
auto_unbox = TRUE,
pretty = TRUE),
json_file)
json_data <- fromJSON(json_file,
simplifyVector = FALSE)
# Open connection to JSONL file
con <- file(jsonl_file, "w")
# Loop over each element (object) and write as one line
for (i in seq_along(json_data)) {
writeLines(toJSON(json_data[[i]], auto_unbox = TRUE), con)
}
# Close connection
close(con)
cat("JSONL saved to:", jsonl_file, "\n")JSONL saved to: /Users/ibeasley/code/genomics_ancest_disease_dispar/output/doccano/abstracts_with_cohort_info.jsonl
sessionInfo()R version 4.3.1 (2023-06-16)
Platform: aarch64-apple-darwin20 (64-bit)
Running under: macOS 15.6.1
Matrix products: default
BLAS: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRlapack.dylib; LAPACK version 3.11.0
locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
time zone: America/Los_Angeles
tzcode source: internal
attached base packages:
[1] stats graphics grDevices datasets utils methods base
other attached packages:
[1] tokenizers_0.3.0 jsonlite_2.0.0 xml2_1.4.0 rentrez_1.2.4
[5] stringi_1.8.7 dplyr_1.1.4 readxl_1.4.5 stringr_1.5.2
[9] workflowr_1.7.1
loaded via a namespace (and not attached):
[1] compiler_4.3.1 renv_1.0.3 promises_1.3.3 tidyselect_1.2.1
[5] Rcpp_1.1.0 git2r_0.36.2 callr_3.7.6 later_1.4.4
[9] jquerylib_0.1.4 yaml_2.3.10 fastmap_1.2.0 here_1.0.1
[13] R6_2.6.1 SnowballC_0.7.1 generics_0.1.4 knitr_1.50
[17] XML_3.99-0.19 tibble_3.3.0 rprojroot_2.1.0 bslib_0.9.0
[21] pillar_1.11.1 rlang_1.1.6 cachem_1.1.0 httpuv_1.6.16
[25] xfun_0.53 getPass_0.2-4 fs_1.6.6 sass_0.4.10
[29] cli_3.6.5 withr_3.0.2 magrittr_2.0.4 ps_1.9.1
[33] digest_0.6.37 processx_3.8.6 rstudioapi_0.17.1 lifecycle_1.0.4
[37] vctrs_0.6.5 data.table_1.17.8 evaluate_1.0.5 glue_1.8.0
[41] whisker_0.4.1 cellranger_1.1.0 rmarkdown_2.30 httr_1.4.7
[45] tools_4.3.1 pkgconfig_2.0.3 htmltools_0.5.8.1