BLUE_estimates_BT function

Description

Estimates individual-level polygenic risk scores (PRS) with uncertainty using a frequentist approach for binary traits. This implementation applies Firth's bias-reduced logistic regression on the discovery sample, computes the coefficient covariance matrix, and uses the delta method to derive PRS variance and confidence intervals.

Usage

BLUE_estimates_BT(
  discovery_pheno,
  discovery_geno_mat,
  target_pheno,
  target_geno_mat,
  significance_level = 0.05,
  max_iterations = 100
)

Arguments

Details

The function fits a Firth logistic regression model using the logistf package to reduce small-sample bias in the discovery set. It extracts SNP effect estimates and their covariance matrix, and propagates this uncertainty through to the individual-level PRS in the target dataset via the delta method. Confidence intervals are derived assuming normality.

Missing or non-estimable coefficients and variances are set to zero.

Value

A data frame with the following columns:

IID

Individual identifier (from the target phenotype file).

PRS

Estimated polygenic risk score for each individual.

Variance

Estimated variance of the PRS.

Lower_Limit

Lower bound of the confidence interval.

Upper_Limit

Upper bound of the confidence interval.

Examples

  bpd <- system.file("Bpd_0_1.txt", package = "iPRSue", mustWork = TRUE)
  bpt <- system.file("Bpt.txt", package = "iPRSue", mustWork = TRUE)
  gd  <- system.file("Gd.txt",  package = "iPRSue", mustWork = TRUE)
  gt  <- system.file("Gt.txt",  package = "iPRSue", mustWork = TRUE)

  results <- BLUE_estimates_BT(
    discovery_pheno    = bpd,
    discovery_geno_mat = gd,
    target_pheno       = bpt,
    target_geno_mat    = gt,
    significance_level = 0.05,
    max_iterations     = 100
  )
  head(results)

BLUE_estimates_QT function

Description

Estimates individual-level polygenic risk scores (PRS) with uncertainty using a frequentist approach for quantitative traits. This implementation fits a multiple linear regression model in the discovery dataset, computes the coefficient covariance matrix, and applies the delta method to propagate uncertainty to the target dataset.

Usage

BLUE_estimates_QT(
  discovery_pheno,
  discovery_geno_mat,
  target_pheno,
  target_geno_mat,
  significance_level = 0.05
)

Arguments

Details

The function fits a multiple linear regression model (lm) using the discovery data. The estimated SNP effects and their covariance matrix are used to compute PRS and associated uncertainty for each individual in the target dataset. Confidence intervals are constructed using the normal approximation.

Missing or non-estimable coefficients and variances are set to zero.

Value

A data frame with the following columns:

IID

Individual identifier (from the target phenotype file).

PRS

Estimated polygenic risk score for each individual.

Variance

Estimated variance of the PRS.

Lower_Limit

Lower bound of the confidence interval.

Upper_Limit

Upper bound of the confidence interval.

Examples

  qpd <- system.file("Qpd.txt", package = "iPRSue", mustWork = TRUE)
  qpt <- system.file("Qpt.txt", package = "iPRSue", mustWork = TRUE)
  gd  <- system.file("Gd.txt",  package = "iPRSue", mustWork = TRUE)
  gt  <- system.file("Gt.txt",  package = "iPRSue", mustWork = TRUE)
 
  results <- BLUE_estimates_QT(
    discovery_pheno    = qpd,
    discovery_geno_mat = gd,
    target_pheno       = qpt,
    target_geno_mat    = gt,
    significance_level = 0.05
  )
  head(results)

GWAS_BT function

Description

Performs genome-wide association study (GWAS) using plink2 logistic model and outputs the GWAS summary statistics with additive SNP effects (beta) and standard errors (se)

Usage

GWAS_BT(plink_path, b_file, discovery_pheno, discovery_cov, thread = 20)

Arguments

Details

The function uses logistic regression to regress the binary phenotype (1/2 coding for controls/cases) on each SNP separately using plink 2. Then the estimated effects and standard errors are adjusted for standardization. It is optional to employ covariates in the model. If no covariate is used, create your covariate file with a constant in the 3rd column (e.g. vector of 1s).

Value

A data frame with two columns:

beta

Estimated effect size (log odds) for each SNP.

se

Standard error of the estimated effect size.

Examples

## Not run: 
  results <- GWAS_BT(
    plink_path = "./plink2",
    b_file = "./binary_file_prefix",
    discovery_pheno = "./discovery_phenotype_file",
    discovery_cov = "./discovery_covariate_file",
    thread = 48
  )
  head(gwas_results)

## End(Not run)

GWAS_QT function

Description

Performs genome-wide association study (GWAS) using plink2 linear model and outputs the GWAS summary statistics with additive SNP effects (beta) and standard errors (se)

Usage

GWAS_QT(plink_path, b_file, discovery_pheno, discovery_cov, thread = 20)

Arguments

Details

The function uses linear regression to regress the quantitative phenotype on each SNP separately using plink 2. Then the estimated effects and standard errors are adjusted for standardization. The phenotype is standardized prior to analysis. It is optional to employ covariates in the model. If no covariate is used, create your covariate file with a constant in the 3rd column (e.g. vector of 1s).

Value

A data frame with two columns:

beta

Estimated effect size from linear regression.

se

Standard error of the effect size estimate.

Examples

## Not run: 
  results <- GWAS_QT(
    plink_path = "./plink2",
    b_file = "./binary_file_prefix",
    discovery_pheno = "./discovery_phenotype_file",
    discovery_cov = "./discovery_covariate_file",
    thread = 48
  )
  head(results)

## End(Not run)

iPRSue_estimates_BT function

Description

Computes individual-level polygenic risk scores (PRS) with uncertainty estimates using a simulation-based approach for binary traits. This implementation follows the iPRSue framework, simulating multiple PRSs by sampling from the GWAS effect size distribution and deriving individual-level confidence intervals.

Usage

iPRSue_estimates_BT(
  gwas,
  target_pheno,
  target_geno_mat,
  no_of_PRSs = 500,
  significance_level = 0.05,
  seed = NULL
)

Arguments

Details

For each SNP, the function simulates no_of_PRSs effect sizes from a normal distribution defined by its GWAS beta and SE. These sampled betas are multiplied by the genotype matrix to generate PRS replicates for each individual. Confidence intervals are then calculated using the specified significance level.

This function is designed for binary traits and should be used with GWAS summary statistics obtained from logistic regression.

Value

A data frame containing the following columns:

IID

Individual identifier (from target phenotype file).

PRS

Mean of simulated PRSs for each individual.

Variance

Variance across simulated PRSs.

Lower_Limit

Lower bound of the confidence interval.

Upper_Limit

Upper bound of the confidence interval.

Examples

## Not run: 
  # Step 1: Run GWAS on binary trait
  results <- GWAS_BT(
    plink_path = "./plink2",
    b_file = "./binary_file_prefix",
    discovery_pheno = "./discovery_phenotype_file",
    discovery_cov = "./discovery_covariate_file",
    thread = 48
  )
  
  # Step 2: Estimate individual PRS with uncertainty
  bpt <- system.file("Bpt.txt", package = "iPRSue", mustWork = TRUE)
  gt  <- system.file("Gt.txt",  package = "iPRSue", mustWork = TRUE)
  
  prs_estimates <- iPRSue_estimates_BT(
    gwas              = results,
    target_pheno      = bpt,
    target_geno_mat   = gt,
    no_of_PRSs        = 500,
    significance_level = 0.05,
    seed              = 123
  )
  head(prs_estimates)

## End(Not run)

iPRSue_estimates_QT function

Description

Computes individual-level polygenic risk scores (PRS) with uncertainty estimates using a simulation-based approach for quantitative traits. This implementation follows the iPRSue framework, simulating multiple PRSs by sampling from the GWAS effect size distribution and deriving individual-level confidence intervals.

Usage

iPRSue_estimates_QT(
  gwas,
  target_pheno,
  target_geno_mat,
  no_of_PRSs = 500,
  significance_level = 0.05,
  seed = NULL
)

Arguments

Details

For each SNP, the function simulates no_of_PRSs effect sizes from a normal distribution defined by its GWAS beta and SE. These sampled betas are multiplied by the genotype matrix to generate PRS replicates for each individual. Confidence intervals are then calculated using the specified significance level.

This function is designed for quantitative traits and should be used with GWAS summary statistics obtained from linear regression.

Value

A data frame containing the following columns:

IID

Individual identifier (from target phenotype file).

PRS

Mean of simulated PRSs for each individual.

Variance

Variance across simulated PRSs.

Lower_Limit

Lower bound of the confidence interval.

Upper_Limit

Upper bound of the confidence interval.

Examples

## Not run: 
  # Step 1: Run GWAS on quantitative trait
  results <- GWAS_QT(
    plink_path = "./plink2",
    b_file = "./binary_file_prefix",
    discovery_pheno = "./discovery_phenotype_file",
    discovery_cov = "./discovery_covariate_file",
    thread = 48
  )
  
  # Step 2: Estimate individual PRS with uncertainty
  qpt <- system.file("Qpt.txt", package = "iPRSue", mustWork = TRUE)
  gt  <- system.file("Gt.txt",  package = "iPRSue", mustWork = TRUE)
  
  prs_estimates <- iPRSue_estimates_QT(
    gwas              = results,
    target_pheno      = qpt,
    target_geno_mat   = gt,
    no_of_PRSs        = 500,
    significance_level = 0.05,
    seed              = 123
  )
  head(prs_estimates)

## End(Not run)