R/pubchem.R

In oboforty/metafetcher: An R package for linking and standardizing metabolites data

library(stringr)

source("R/db_ctx.R")
source("R/utils.R")


create_pubchem_record <- function () {
  pubchem_attribs <- c(
    "smiles", "inchi", "inchikey", "formula", "names",
    "mass", "monoisotopic_mass", "logp",

    "pubchem_id", "chebi_id",  "hmdb_id", "kegg_id",
    "ref_etc"
  )

  df <- data.frame(matrix(ncol = length(pubchem_attribs), nrow = 1))
  colnames(df) <- pubchem_attribs

  # vector fields: names, smiles
  df$names <- list(vector(length=0))
  df$smiles <- list(vector(length=0))

  return(df)
}


PubchemHandler <- setRefClass(Class = "PubchemHandler",
  fields = list(
    name = "character"
  ),
  methods = list(
    initialize=function(...) {
      callSuper(...)
        # Initialise fields here (place holder)...
        .self
    },

    query_metabolite = function(db_id) {
      # Queries a KEGG metabolite record and converts it to a common interface
      SQL <- "SELECT
        pubchem_id, chebi_id, kegg_id, hmdb_id,
        smiles, inchi, inchikey, formula, names,
        mass, monoisotopic_mass, logp
        FROM pubchem_data WHERE pubchem_id = '%s'"
      df.pubchem <- db.query(sprintf(SQL, db_id))

      if (length(df.pubchem) == 0) {
        # call api and save to database
        df.pubchem <- .self$call_api(db_id)

        # if api response is still empty, then the record doesn't exist
        if (is.null(df.pubchem) || length(df.pubchem) == 0)
          return(NULL)
        #
        # # Save to db
        # if (length(df.pubchem$names[[1]]) > 0)
        #   df.pubchem$names <- c(join_sql_arr(df.pubchem$names[[1]]))
        #
        # if (length(df.pubchem$smiles[[1]]) > 0)
        #   df.pubchem$smiles <- c(join_sql_arr(df.pubchem$smiles[[1]]))

        # cache pubchem record
        db.write_df("pubchem_data", convert_df_to_db_array(df.pubchem, c("names", "smiles")))
      } else {
        # we don't have to convert from postgres array format if the data comes from api

        # convert pg array strings to R vectors:
        df.pubchem$names <- list(pg_str2vector(df.pubchem$names[[1]]))
        df.pubchem$smiles <- list(pg_str2vector(df.pubchem$smiles[[1]]))
      }

      return (df.pubchem)
    },

    query_reverse = function(df.res) {
      chebi_id <- df.res$chebi_id[[1]]
      kegg_id <- df.res$kegg_id[[1]]
      hmdb_id <- df.res$hmdb_id[[1]]

      # construct complex reverse query
      SQL <- "SELECT pubchem_id FROM pubchem_data WHERE"
      clauses <- character()

      # todo: itt: construct proper is empty!
      if (!is.empty(chebi_id))
        clauses <- c(clauses, sprintf("chebi_id = '%s'", chebi_id))
      if (!is.empty(hmdb_id))
        clauses <- c(clauses, sprintf("hmdb_id = '%s'", hmdb_id))
      if (!is.empty(kegg_id))
        clauses <- c(clauses, sprintf("kegg_id = '%s'", kegg_id))

      if (length(clauses) == 0)
        return(NULL)

      SQL <- paste(SQL, paste(clauses, collapse = " OR "))
      df.lipidmaps <- db.query(SQL)

      if(length(df.lipidmaps) == 0) {
        return(NULL)
      }

      return(df.lipidmaps$chebi_id)
    },

    call_api = function(db_id) {
      'Calls PubChem api to retrieve record.'
      df.pubchem <- create_pubchem_record()

      # 1. Parse properties
      url <- 'https://pubchem.ncbi.nlm.nih.gov/rest/pug/compound/cid/%s/json'
      v <- http_call_api(url, db_id)

      if (is.null(v))
        return(NULL)

      df.pubchem$pubchem_id <- v$PC_Compounds[[1]]$id$id$cid
      props <- v$PC_Compounds[[1]]$props

      # todo: multiple cardinality?
      for (prop in props) {
        label <- prop$urn$label
        val <- prop$value

        if (label == 'InChI')
            df.pubchem$inchi <- lstrip(val$sval, "InChI=")
        else if (label == 'InChIKey')
            df.pubchem$inchikey <- val$sval
        else if (label == 'SMILES')
            df.pubchem$smiles[[1]] <- c(df.pubchem$smiles[[1]], val$sval)
        else if (label == 'IUPAC Name')
            df.pubchem$names[[1]] <- c(df.pubchem$names[[1]], val$sval)
        else if (label == 'Molecular Formula')
            df.pubchem$formula <- val$sval
        # else if (label == 'Mass')
        #     df.pubchem$mass <- val$fval
        else if (label == 'Molecular Weight')
            df.pubchem$mass <- val$fval
        else if (label == 'Weight' && prop$urn$name == 'MonoIsotopic')
            df.pubchem$monoisotopic_mass <- val$fval
        else if (label == 'Log P')
            df.pubchem$logp <- val$fval
      }

      # 2. parse external references
      url <- 'https://pubchem.ncbi.nlm.nih.gov/rest/pug/compound/cid/%s/xrefs/SourceName,RegistryID/JSON'
      v <- http_call_api(url, db_id)

      if (is.null(v))
        return(NULL)

      ids <- v$InformationList$Information[[1]]$RegistryID

      for (xdb_id in ids) {
        # todo: rest, e.g. chemspider?

        if (startsWith(xdb_id, 'CHEBI:'))
          df.pubchem$chebi_id <- lstrip(xdb_id, "CHEBI:")
        else if (startsWith(xdb_id, 'HMDB'))
          df.pubchem$hmdb_id <- xdb_id


      #  else if (substr(xdb_id, 1, 1) == 'C' && str_detect(xdb_id, '^C\\d{4,9}$'))
        else if (substr(xdb_id, 1, 1) == 'C' && grepl('^C\\d{4,9}$',xdb_id))
          df.pubchem$kegg_id <- xdb_id
      }
      #print(df.pubchem)

      return(df.pubchem)
    }

  )
)