HAPNEST: efficient, large-scale generation and evaluation of synthetic datasets for genotypes and phenotypes

Area of Science:

• Genomics
• Computational Biology
• Statistical Genetics

Background:

• Existing methods for simulating synthetic genotype and phenotype data lack scalability for large-scale analyses.
• There is a need for systematic evaluation of synthetic data quality and benchmark datasets for polygenic risk score development.

Purpose of the Study:

• To introduce HAPNEST, an efficient approach for generating diverse individual-level genotypic and phenotypic data.
• To demonstrate HAPNEST's utility in facilitating biobank-scale analyses and evaluating polygenic risk score methods.

Main Methods:

• HAPNEST was developed for efficient generation of synthetic genotypic and phenotypic data.
• The approach was used to create a large dataset of 1 million individuals, 6.8 million variants, and nine phenotypes.
• Seven polygenic risk score generation methods were compared using the synthetic data across different ancestries and genetic architectures.

Main Results:

• HAPNEST offers faster computational speed and lower relatedness to reference panels compared to alternatives.
• The generated synthetic data preserves key statistical properties of real-world data.
• A large-scale synthetic dataset (1,008,000 individuals, 6.8M variants, 9 traits) was successfully generated.

Conclusions:

• HAPNEST provides a scalable and efficient solution for generating high-quality synthetic biological data.
• The developed synthetic dataset and HAPNEST software facilitate large-scale genomic research and polygenic risk score development.