Find the top 10 correlated genes, based on data from ARCHS4

find_correlated_genes.R

# find_correlated_genes.R
# 2024-03-12
# Kamil Slowikowski

library(arrow) # install.packages("arrow")

# Download correlation from ARCHS4 <https://maayanlab.cloud/archs4/download.html>
# wget https://s3.amazonaws.com/mssm-data/human_correlation_archs4.f

# The file is large:
# > utils:::format.object_size(file.size("human_correlation_archs4.f"), "auto")
# [1] "5.2 Gb"

# Memory-map the data into our R session:
d <- read_feather("human_correlation_archs4.f")

# We read data lazily, as-needed:
# > pryr::object_size(d)
# 6.40 MB

# Here is a simple function to find the top 10 correlated genes:
correlated_genes <- function(my_gene = "CXCL8", n = 10) {
  stopifnot(my_gene %in% colnames(d))
  i <- which(colnames(d) == my_gene)
  x <- d[,i]
  o <- order(-x)
  values <- x[o]
  names(values) <- colnames(d)[o]
  head(values, n = n)
}

# Usage examples:

# > correlated_genes("IL6")
#       IL6      CSF3       TNF   ZC3H12A      IL1B     PTGS2     ICAM1      IL27     ACOD1      IL1A
# 1.0000000 0.5144461 0.4947951 0.4922328 0.4828641 0.4776033 0.4670185 0.4648258 0.4582155 0.4370421
#
# > correlated_genes("GRN")
#       GRN      CTSZ      CD68    MAN2B1     FCGRT   SLC15A3      CTSD     IFI30      PSAP     MGAT1
# 1.0000000 0.7919509 0.7654468 0.7480757 0.7458239 0.7423101 0.7395988 0.7245633 0.7099952 0.6936354
#
# > correlated_genes("PLAC8")
#     PLAC8      PLD4      IRF8     VAMP8      RAC2       SCT     MS4A3      GMFG      LSP1     PSME1
# 1.0000000 0.4843658 0.4672987 0.4208228 0.3617905 0.3185449 0.3137622 0.3133707 0.3121387 0.3034305