train_and_impute_PRS {gnonadd}

Trains and imputes a poligenic risk score (PRS)

Description

This function trains a poligenic risk score model on a dataset, and then imputes the risk score into another dataset

Usage

train_and_impute_PRS(
  qt_training,
  g_training,
  g_impute,
  dominance_effects = rep(FALSE, ncol(g_training)),
  interaction_effects = matrix(0, nrow = 0, ncol = 0)
)

Arguments

qt_training

A numeric vector. Represents the qt values of the data we train the model on.

g_training

A matrix, where each colomn represents a variant and each line represents a subject in the training data

g_impute

A matrix, where each column represents a variant and each line represents a subject.

dominance_effects

A Boolian vector, each term determines whether a dominance term for the corresponding variant is used in the model.

interaction_effects

An integer matrix with two columns. Each line represents a pair of interacting variants that should be included in the model.

Value

Returns a list with the following objects * PRS_imputed, the imputed PRS values * PRS_training, the PRS values for the training data * Residuals_training, the residuals from the model in the training data

Examples

g_train_vec <- matrix(0, nrow = 100000, ncol = 5)
freqs <- runif(ncol(g_train_vec), min = 0, max = 1)
for(i in 1:ncol(g_train_vec)){
 g_train_vec[,i] <- rbinom(100000, 2, freqs[i])
}
g_impute_vec <- matrix(0, nrow = 50000, ncol = 5)
for(i in 1:ncol(g_impute_vec)){
 g_impute_vec[,i] <- rbinom(50000, 2, freqs[i])
}
dom_vec <- c(TRUE, FALSE, FALSE, TRUE, FALSE)

int_vec <- matrix(c(1, 2, 4, 5), nrow = 2 , ncol = 2)

qt_vec <- rnorm(100000) + 0.2 * g_train_vec[, 1] + 0.3 * g_train_vec[, 1] * g_train_vec[, 4]

res <- train_and_impute_PRS(qt_vec, g_train_vec, g_impute_vec,
       dominance_effects = dom_vec, interaction_effects = int_vec)