bigsnpr R Package

The core data structure in bigsnpr is the bigSNP class, which represents SNP data as a list containing:

• $genotypes: An FBM.code256 object where rows are samples and columns are variants. This stores genotype calls or dosages (rounded to 2 decimal places).
• $fam: A data.frame containing individual/sample information.
• $map: A data.frame containing variant information.

Important Note on Missing Values: Most algorithms in this package do not handle missing values. You should either impute missing values of genotyped variants using snp_fastImpute() or snp_fastImputeSimple(), or use the bed_*() functions which are designed to work directly on .bed files with a few missing values.

The package provides three primary methods for polygenic score calculation:

1. Penalized regressions: Uses individual-level data.
2. Clumping and Thresholding (C+T) and Stacked C+T (SCT): Uses a combination of summary statistics and individual-level data.
3. LDpred2 and lassosum2: Uses summary statistics (LDpred2 is a major focus of the package).

You can install bigsnpr from CRAN or directly from GitHub using the remotes package.

From CRAN:

install.packages("bigsnpr")

From GitHub:

# install.packages("remotes")
remotes::install_github("privefl/bigsnpr")

Use the following functions to load genotype data into the bigSNP format:

• PLINK files (bed/bim/fam): Use snp_readBed() or snp_readBed2().
• UK Biobank BGEN files: Use snp_readBGEN().

Before reading into the bigSNP format, you can perform quality control and conversion using PLINK via the R functions snp_plinkQC() and snp_plinkKINGQC().

# Example usage (conceptual):
# bfile <- snp_readBed(bedfile = "path/to/file")
# bgen <- snp_readBGEN(bgenfile = "path/to/file")

Since most bigsnpr algorithms require complete data, you can use these functions to handle missing values in genotyped variants:

• snp_fastImpute(): A more thorough imputation method (can take several hours for large datasets, e.g., 15K x 300K).
• snp_fastImputeSimple(): A faster, simpler imputation method (takes only a few minutes).