mhcnpreds: mhcnpreds: Lookup tables for MHCnuggets IC50 predictions
In richelbilderbeek/mhcnpreds: MHCnuggets Predictions

Description Details Author(s) Examples

We can predict how well a peptide binds to the presentation molecules used by the immune system. This package allows to see how well a molecule binds relatively.

The mhcnuggetsr R package can predict the IC50 (50 concentration) of a peptide for an MHC (I or II) haplotype. A peptide that is strongly bound to the MHC molecules is in the lower percentile, for example in the lowest 2 This package determines the threshold IC50 for all percentiles.

To be able to do so, the IC50 predictions of 10k random peptides is measured and the distribution at the percentile level is stored as a lookup table.

Richèl J.C. Bilderbeek

if (mhcnuggetsr::is_mhcnuggets_installed()) {
  # mhcnuggetsr has a list of MHC haplotypes
  mhc_haplotype <- mhcnuggetsr::get_trained_mhc_1_haplotypes()[100]
} else {
  # or use a hard-coded haplotype, in MHCnuggets naming format
  mhc_haplotype <- "HLA-B83:01"

}

# Get the IC50 value of which 2% of IC50 predictions are below,
# that is, bind better to the MHC complexes
ic50_threshold <- get_ic50_threshold(
  peptide_length = 9,
  mhc_haplotype = mhc_haplotype,
  percentile = 0.02
)

# A IC50 value predicted by MHCnuggets
my_ic50 <- 100.0

if (my_ic50 < ic50_threshold) {
  message("This is a binder, as it is in the lowest 2% of all IC50s")
}