PANTHER FUNCTIONAL CLASSIFICATION in R

Question

Does anyone know of a way to programmatically get panther protein class terms and IDs for a gene list, in a way that the classification structure is preserved (PARENT CLASS + CHILD CLASS)?

i.e. For a given gene, I would want to get the list of classes, in away that I can then use the top classification for downstream analyses.

GENEX, PC00244 zinc finger transcription factor > PC00218 DNA-binding transcription factor > PC00264 gene-specific transcriptional regulator

Answer 1

One approach is to get the gene's sequence (DNA/AA) and use the interproscan API (e.g., https://www.ebi.ac.uk/Tools/common/tools/help/index.html?tool=iprscan5) with the httr2 library in R to POST gene sequence(s) and GET the resulting Panther annotation. Make sure to include Panther in the appl parameter for the post request.

Here's a function that does this for a single sequence.

submit_fasta <- function(
    fasta,
    path_outfolder,
    email,
    stype = "n",
    appl = c("Panther"),
    title = 'stackoverflow'
    ) {
  stopifnot("stype must be 'n' for nucleotide seqs or 'p' for protein seqs" = stype %in% c('n', 'p'))
  url_post <- "https://www.ebi.ac.uk/Tools/services/rest/iprscan5/run"
  chr_seq <- as.character(fasta)
  req <- httr2::request(url_post)
  req <- req |> httr2::req_body_form(
    "email" = email,
    "title" = title,
    "goterms" = "false",
    "pathways" = "false",
    "stype" = stype,
    "appl" = paste0(appl, collapse = ","),
    "sequence" = chr_seq
  )
  req <- req |> httr2::req_headers("Accept" = "text/plain", "Content-type" = "application/x-www-form-urlencoded")
  resp_post <- req |> httr2::req_perform()
  chr_retrieval_code <- resp_post$body |> rawToChar()
  url_get <- "https://www.ebi.ac.uk/Tools/services/rest/iprscan5/result"
  req <- httr2::request(url_get)
  req <- req |> httr2::req_url_path_append(chr_retrieval_code)
  req <- req |> httr2::req_url_path_append("tsv")
  req <- req |> httr2::req_retry(
    max_tries = 15L, backoff = ~ 30L,
    is_transient = function(resp) {httr2::resp_status(resp) == 400L}
  )
  req <- req |> httr2::req_headers(Accept = NULL)
  req |> httr2::req_dry_run(quiet = FALSE)
  resp_get <- req |> httr2::req_perform()
  cat('status code:', resp_get$status_code, '\n')
  result <- resp_get$body |> rawToChar()
  result <- read.table(text = result, sep = '\t')
  return(result)
}

Importantly, the fasta input parameter is expected to be XString object from the Biostrings package. However, the code could easily be adjusted. The sequence eventually needs to be coerced to a string for the POST request.

#fasta <- Biostrings::readDNAStringSet('seq.fa')
#resp <- submit_fasta(fasta = fasta_single, path_outfolder = 'x', email = '[email protected]', title = 'stackoverflow')

edit: set the is_transient param to handle the actual transient error case where the data is not yet available (response code = 400). Also, it's important to note there's a "status" endpoint https://www.ebi.ac.uk/Tools/common/tools/help/index.html?tool=iprscan5#/Status) which is probably much more appropriate than pinging the data endpoint until either success or reaching max tries. However, I hope this answer suffices!