Reading input data

MiDAS includes functions to read HLA and KIR typing data, checking for the correct format and adherence to the right nomenclature. HLA calls can be reduced to a desired resolution when importing it, e.g. from 6-digit to 4-digit. Phenotypic observations can be read e.g. using the read.table function (make sure to include stringsAsFactors = FALSE argument).

pheno_file <- system.file("extdata", "MiDAS_tut_pheno.txt", package = "midasHLA")
pheno <- read.table(pheno_file, header = TRUE, stringsAsFactors = FALSE)

# HLA calls can be loaded using the readHlaCalls function with the desired resolution
hla_calls_file <- system.file("extdata", "MiDAS_tut_HLA.txt", package = "midasHLA")
hla_calls <- readHlaCalls(hla_calls_file, resolution = 4)

# KIR calls (currently presence/absence calls, no allele-level resolution) can be loaded using the readKirCalls function
kir_calls_file <- system.file("extdata", "MiDAS_tut_KIR.txt", package = "midasHLA")
kir_calls <- readKirCalls(kir_calls_file)

For convenience, we will here use test data shipped with the package.

HLA <- reduceHlaCalls(MiDAS_tut_HLA, resolution = 4)
KIR <- MiDAS_tut_KIR
pheno <- MiDAS_tut_pheno

Creating midas objects

MiDAS provides the prepareMiDAS function to combine genetic and phenotypic data for subsequent analysis. In this step, it is also possible to infer amino acid level information from HLA calls, encode HLA-KIR interactions, or calculate HLA evolutionary divergence (experiment argument).

midas <- prepareMiDAS(
  hla_calls = HLA,
  kir_calls = KIR,
  colData = pheno,
  experiment = c("hla_alleles", "hla_aa")
  )

MiDAS object

A MiDAS object contains our input data as well as defined transformations. This is an extension of a MultiAssayExperiment and can be handled as such. MiDAS provides some functions to interact with MiDAS objects (check ?MiDAS-class for a full list).

For example, we can explore the HLA alleles frequencies, compare them with published frequencies from the allelefrequencies.net database, and filter out low frequency alleles.

freq <- getFrequencies(
  object = midas,
  carrier_frequency = FALSE,
  experiment = "hla_alleles",
  compare = TRUE
)

midas <-
  filterByFrequency(
    object = midas,
    experiment = "hla_alleles",
    lower_frequency_cutoff = 0.01
    )

Association analysis

# Logistic regression
object <- glm(disease ~ term, data = midas, family = binomial(link = "logit"))

results <-
  runMiDAS(
    object = object,
    experiment = "hla_alleles",
    inheritance_model = "dominant",
    conditional = FALSE,
    omnibus = FALSE,
    lower_frequency_cutoff = 0.05
    )

kableResults(results)