File to use in support ticket

stringdb_combined_score.R

library(tidyverse)
library(STRINGdb)

STRINGdb::get_STRING_species(version = version, species_name=NULL) %>%
    dplyr::arrange(official_name) %>%
    dplyr::filter(official_name == 'homo_sapiens')

# downloading Homo sapiens
string_db <- STRINGdb::STRINGdb$new(version         = version,
                                    species         = 9606,
                                    score_threshold = 0)

# Load to memory the database (by calling a method)
tp53 <- string_db$mp( "tp53" )

# get all interactions
all.interactions <- as.tbl(string_db$get_interactions(string_db$proteins$protein_external_id))

col.ixs <- (!colnames(all.interactions) %in% 'combined_score')

# columns without from and to
col.vals.ixs <- col.ixs & (!colnames(all.interactions) %in% c('from', 'to'))

p <- 0.041

mat <- as.matrix(all.interactions[,col.vals.ixs]) / 1000

#
# Calculate new combined score
#  https://string-db.org/help/faq/#how-are-the-scores-computed

combined.score <- mat %>%
  apply( 2, function(ix) {
    res <- (ix - p) / (1 - p)
    res[ix == 0] <- 0
    return(res)
  }) %>% {
    # Normalize Co-Occurence and Text mining scores
    #  https://string-db.org/help/faq/#how-are-the-scores-computed

    for (ix in c('cooccurence', 'textmining', 'textmining_transferred')) {
      if (ix %in% colnames(.)) {
        .[, ix] <- .[, ix] * (1 - .[, 'homology'])
      }
    }
    .[,!colnames(.) %in% c('homology')]
    #.
  } %>% {
    res <- (1 - .[,1])
    for (ix in seq(ncol(.))[-1]) {
      res <- res * (1 - .[,ix])
    }
    1 - res
  } %>% {
    . + p * (1 - .)
  } %>% {
    floor(. * 1000)
  }

abs(combined.score - all.interactions$combined_score) %>%
  data.frame() %>% filter(. > 1) %>% { qplot(.[,'.'], bins = 100) }