GPTL: Genomic Prediction Using Transfer Learning

GPTL is an R package that implements several methods for Genomic Prediction using Transfer Learning, including

• TL-GDES: A Gradient Descent Algorithm with Early Stopping.
• TL-PR: A Penalized regression with shrinkage towards a prior mean (e.g., estimates derived from another data set).
• TL-BMM: A Bayesian model with a finite mixture prior that allows TL from multiple prior sources of information.

Citation

Please cite Wu et.al for the GPTL package.

Installation

• Install the package from GitHub

# install.packages("remotes")
remotes::install_github("QuantGen/GPTL")

Using GPTL

GPTL offers three polygenic score methods using Transfer Learning (TL). The functions GD(), PR(), and BMM() implement Gradient Descent with Early Stopping, Penalized Regression, and Bayesian Mixture model, respectively.

These functions take as input SNP effects estimates from a source population (used as prior values to the TL algorithm) and sufficient (or summary) statistics from the target population. The sufficient statistics (𝑿′𝑿, 𝑿′𝒚) can be computed from individual genotype-phenotype data or reconstructed from GWAS results and an LD reference panel.

We provide below two links to human LD reference panels derived using All of Us and UK-Biobank, and examples illustrating how to use each of the functions included in GPTL.

Human LD reference panels

LD reference panels constructed using the All of Us data (CDRv7, Controlled Tier)

• African American (~3.5G)

  wget https://zenodo.org/records/17686189/files/AA_AOU.tar.gz
  tar -zxvf AA_AOU.tar.gz

• Hispanic (~2.7G)

  wget https://zenodo.org/records/17686189/files/HIS_AOU.tar.gz
  tar -zxvf HIS_AOU.tar.gz

LD reference panels constructed using the UK Biobank data

• African (~5.1G)

  wget https://zenodo.org/records/17686189/files/AFR_UKB.tar.gz
  tar -zxvf AFR_UKB.tar.gz

• American (~4.3G)

  wget https://zenodo.org/records/17686189/files/AMR_UKB.tar.gz
  tar -zxvf AMR_UKB.tar.gz

Examples

• 1. GPTL using individual genotype-phenotype data
• 2. GPTL using LD reference panel and GWAS results

System Requirements

• Depends: R (>= 3.5.0)
• This package is compatible with Windows, Mac, and Linux operating systems and has been tested on Windows 7 & 10, macOS Sequoia & Sonoma, and Linux CentOS 7.

Support

Please direct any problems or questions to Hao Wu ([email protected]) or Gustavo de los Campos ([email protected]).