I'm trying to use this europmc r libray where I have a list of pmids to look for. I tried with pubtator but its bit complicated.In Europmc i can all the annotated terms etc.


example list of PMIDS

Now what I'm doing is looking each ID using epmc_details function which is not the way i would do if i have to look for hundreds.

epmc_details(ext_id = '30510081')

I have to question how can I run the epmc_details in a loop or some other way where i can look for PMIDS one by one and save the result in a data frame.

The epmc_details is returned as a list. The structure of the list is as such

[1] "basic"           "author_details"  "journal_info"    "ftx"             "chemical"        "mesh_topic"      "mesh_qualifiers"
[8] "comments"        "grants"    

I would only like to save basic,chemical,mesh_topic,mesh_qualifiers in one data frame.

For example if my first id is this 30510081 the dataframe should have first column as my ID which is basically basic[1]and rest of the information appended to the next columns.

such as

 ID chemical mesh_qualifiers mesh_topic gene 

Any suggestion or help would be really and highly appreciated

I was looking at europmc site through the browser

this was one of my query

when i highlight the keyterms i do see in the abstract itself all the keyterms are getting annotated but when i do the same query search through R I do see empty results as such why there is a difference?

# A tibble: 0 x 0

# A tibble: 0 x 0

# A tibble: 0 x 0

I found better way of getting data from pubmed using the tidypmc library.

doc <- pmc_xml("PMC6365492")
txt <- pmc_text(doc)

count(txt, "section")

cap1 <- pmc_caption(doc)
filter(cap1, sentence == 1)

tab1 <- pmc_table(doc)

sapply(tab1, nrow)


collapse_rows(tab1, na.string="-")

x <- xml_name(xml_find_all(doc, "//*"))
tibble(tag=x) %>% count("tag")

x1 <- unnest_tokens(txt, word, text) %>%
  anti_join(stop_words) %>%
  filter(!word %in% 1:100)
#  Joining, by = "word"
#filter(x1, str_detect(section, "Case description")) 

filter(x1, str_detect(section, "Results"))

tbls <- pmc_table(doc)
map_int(tbls, nrow)


collapse_rows(tbls, na.string="-")

But if i understand it can use one PMC id at a time. Again keeping my original question how can i put this in a loop to query lets say i have 100 PMCID and get it result and store in a dataframe.

After using tidypmc i found that i parse all the publication based on attributes or tags. Such as title,abstract, results etc etc.

Lets say I'm interested in the table tags information where they have metadata of patients as well as others.So if a paper contain multiple tables I would like to store each of them in a data-frame under the respective publication.Since I have multiple IDs to search and do the save as above mentioned. How to put this through a loop or can it be done without using loop ?

Any suggestion or help would be really appreciated as always.

  • 1
    $\begingroup$ I think the content of that article are not available in pubmed, you can only download a pdf through some link.. again.. it's my guess.. maybe if you do more you will see this repeats itself $\endgroup$ – StupidWolf Jul 26 '20 at 8:55
  • $\begingroup$ @StupidWolf i found tidypmc is better but yes it will only look for article which has PMCID . I have updated my question and query, I would be glad if you can have a look and give your suggestion. $\endgroup$ – krushnach Chandra Jul 27 '20 at 10:56

The idea for this code is to first convert PIDs to PMCIDs, then run the tidypmc in a loop over the PMCIDs. The only problem is that tidypmc failed to retrieve tables from most of the IDs in your example list.

example_pids <- c(30024784, 30555165, 30510081, 31688884, 31516032, 28588019, 29286103) %>% as.character()

#-- Convert to PMC ids
convertPIDtoPMCID <- function(pids) {
    #-------- Make API request
    pids4query <- paste(pids, collapse = "%0D%0A")
    idconv_req <- paste0("https://www.ncbi.nlm.nih.gov/pmc/utils/idconv/v1.0/?ids=", pids4query, "&idtype=pmid&format=json&versions=no&showaiid=no&tool=&email=&.submit=Submit")
    pids_json <- GET(idconv_req) %>% content("text") %>% fromJSON()
    #-------- Get info from JSON
    pid2pmc <- pids_json$records %>% 
        select(pmcid, pmid) %>% 
    rownames(pid2pmc) <- pid2pmc$pmid
    pmcids <- pmcids <- pid2pmc[pids, "pmcid"]
example_pmcids <- convertPIDtoPMCID(example_pids)

#-- Try to get data with tidypmc
pub_tables <- lapply(example_pmcids, function(pmc_id) {
    message("-- Trying ", pmc_id, "...")
    doc <- tryCatch(pmc_xml(pmc_id), 
                    error = function(e) {
                        message("------ Failed to recover PMCID")
    if(!is.null(doc)) { 
        #-- If succeed, try to get table
        tables <- pmc_table(doc)
        if(!is.null(tables)) {
            #-- If succeed, try to get table name
            table_caps <- pmc_caption(doc) %>%
                filter(tag == "table")
            names(tables) <- paste(table_caps$label, table_caps$text, sep = " - ")
    } else {
        #-- If fail, return NA
names(pub_tables) <- example_pids

#-- Inspect results
pub_tables$`30555165`$`Table 1 - Patient demographic and baseline characteristics`
pub_tables$`29286103`$`Table I - Sample summary.`

Tables will require quite a bit of tidying after this. Good luck!

  • $\begingroup$ github.com/ropensci/tidypmc/issues/3 this answer from the library developer $\endgroup$ – krushnach Chandra Jul 27 '20 at 19:20
  • $\begingroup$ b <-epmc_search(query = 'acute myeloid leukemia drug studies',output = 'parsed',limit = 200) a <- b %>% select(pmid,pmcid) a <- a[complete.cases(a),] c <- a$pmcid Error in names(tables) <- paste(table_caps$label, table_caps$text, sep = " - ") : 'names' attribute [4] must be the same length as the vector [2] Im getting this error when i run the pub_tables chunk $\endgroup$ – krushnach Chandra Jul 27 '20 at 21:29
  • 1
    $\begingroup$ That line is just for embellishment, you shouldn't have a problem removing it. It adds the title of the table. But it might fail if the metadata doesn't have the table title. So feel free to remove it. $\endgroup$ – csgroen Jul 28 '20 at 9:48
  • 1
    $\begingroup$ Most publications with a Pubmed ID are not open access and you cannot mine the text, so be sure to filter results using OPEN_ACCESS:Y in Europe PMC $\endgroup$ – Chris S. Jul 28 '20 at 15:44
  • 1
    $\begingroup$ @krushnachChandra Tabular data represents m observations of n variables. The issue here is that each table has different variables and observations. How can you join tables that have different columns and different rows? You can't. That's why we get a list. $\endgroup$ – csgroen Jul 30 '20 at 8:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.