Population

Sample size

Phenotype

Variants

VCF files


GWAS summary


Top associated variants

The table shows up to the top 100 variants sorted by P-value. Select a row to open that variant directly in the Locus zoom tab.

Manhattan plot (interactive)

Click any point to open a detailed regional ("locus zoom") view in the Locus zoom tab.

QQ plot
Regional association plot

Point color reflects LD (r²) with the reference SNP, when available in the LD file for this chromosome. The clicked SNP is a purple diamond; if it isn't part of the thinned LD reference set, the nearest available reference SNP is shown as a red star instead.

Dataset evaluation

Evaluation of genetic properties, analytical utility, and privacy-related properties of the selected synthetic dataset.


PCA, LD, and MIA results are chromosome-specific. Kinship is evaluated across the dataset and is independent of this selection.


Population structure

Reference and synthetic samples projected into the same PCA space using principal component loadings derived from the reference dataset.

Linkage disequilibrium

Local LD patterns for the selected chromosome, based on pairwise r² values calculated on matched variant sets.

Download dataset files

Downloads follow the population, sample size and phenotype selected in the sidebar.


Genotypes

Synthetic genotype files are distributed via SFTP. Select one or more chromosomes to download.

The VCF files themselves are not transferred through this web portal. The portal generates a small SFTP script that you can run from your local computer or from a compute environment with SFTP access.
1. Choose chromosomes
2. Download the SFTP script

The script contains the SFTP commands needed to retrieve the selected chromosome-level VCF files.

3. Run the script

Open a terminal in the directory where you want to save the genotype files and run:

The generated script automatically follows the population, sample size and chromosome selection above.

4. Verify file integrity

After downloading the genotype files, place the checksum file in the same directory and verify the files with:

The checksum file contains only the chromosomes selected above. Successfully verified files are reported as OK.

Phenotypes

Download the individual-level synthetic phenotype file for the selected trait.

GWAS summary statistics

Download the complete summary-statistics file for the selected phenotype.

Causal variants

Download the list of causal variants used to simulate the selected phenotype for the selected population.

The causal-variant list depends on population and phenotype and is shared across cohort sizes.

Simulation configuration

Download the YAML configuration file used to generate the selected synthetic dataset.

The configuration corresponds to the population and sample size selected in the sidebar.

Variant ID mapping

Mapping between GRCh38 variant identifiers used in the VCF files (chr:position:REF:ALT) and corresponding rsIDs.

Complete mapping

Complete mapping across chromosomes 1-22.

Evaluation files

Download chromosome-specific PCA, LD results, and privacy-evaluation plots when available. Kinship is dataset-level and independent of the selected chromosome.

Global evaluation

Global assessment of genetic fidelity, causal-variant recovery and computational scalability across populations and cohort sizes.

Allele-frequency similarity and population structure are shown for chromosome 21 as a representative chromosome, while causal-variant recovery is based on genome-wide GWAS results.


Genetic fidelity

These analyses evaluate how closely the synthetic genotype datasets reproduce genetic properties of the reference data across populations and cohort sizes.

Minor allele frequency similarity

MAF distance between synthetic and reference datasets. Lower values indicate greater similarity in allele-frequency distributions.

Global population structure

Combined PCA visualization across populations and cohort sizes, providing an overview of how synthetic samples reproduce reference population structure.

Causal variant recovery

The proportion of predefined causal variants represented in the synthetic genotype datasets was evaluated across populations and cohort sizes.

As cohort size increases, a greater proportion of causal variants is expected to be represented.

Recovered causal variants
Computational scalability

Computational performance was evaluated across increasing cohort sizes to characterize the scalability of the simulation workflow.

Running time

Total runtime required to generate synthetic datasets across populations and cohort sizes.

Peak memory usage

Maximum memory usage observed during synthetic dataset generation across populations and cohort sizes.
Synthetic genomic datasets

Synthetic genotype-phenotype datasets provide a flexible and privacy-conscious resource for human genetics research, enabling data sharing, method development, and large-scale computational studies without relying on access to sensitive individual-level genomic data.

This portal provides an open-access collection of synthetic genotype-phenotype datasets generated using an extended simulation framework based on HAPNEST.

The resource spans multiple ancestry groups, cohort sizes ranging from 1,000 to 2 million individuals, and ten simulated phenotypes.

  • Populations: EUR, EAS, CSA, AFR, AMR
  • Sample sizes: 1K, 10K, 100K, 1M and 2M individuals
  • Ten simulated phenotypes
  • Chromosome-level genotype datasets distributed via SFTP
  • GWAS summary statistics
  • Trait-specific causal variant lists
  • Population- and cohort-specific simulation configuration files
  • Interactive Manhattan plot and QQ plot
  • Global evaluation of genetic fidelity, analytical utility, privacy-related properties, and computational scalability

How these data were generated
Genotypes

Synthetic genotypes were generated using an extended simulation framework based on HAPNEST (Wharrie et al., 2023) , which generates synthetic haplotypes by resampling and recombining genomic segments from reference haplotypes. The reference panel combines individuals from the 1000 Genomes Project and the Human Genome Diversity Project (HGDP) and was filtered to retain HapMap3 variants. This population-diverse reference panel enables the simulation of ancestry-specific genomic structure, allele-frequency patterns, and linkage disequilibrium across the simulated ancestry groups. The resulting synthetic datasets contain 1,329,052 variants across the autosomal chromosomes and share the same variant set as the pre-processed reference data by construction.

Datasets were generated for five ancestry groups (EUR, EAS, CSA, AFR, and AMR) and across a range of cohort sizes, from 1,000 to 2 million individuals.

Phenotypes

Ten phenotypes were simulated for each synthetic cohort: height, BMI, HDL cholesterol, diastolic blood pressure, triglyceride levels, body weight, resting heart rate, alcohol consumption, HbA1c, and lymphocyte count.

For each phenotype and ancestry group, a predefined set of causal variants was used to generate the genetic component of the simulated trait. Phenotype simulation parameters, including the settings used for genotype and phenotype generation, are recorded in the corresponding YAML configuration file.

Data availability and reproducibility

To support transparency and reproducibility, the resources associated with each dataset are available through the Download tab. Users can retrieve synthetic genotype datasets via SFTP and directly download individual-level phenotype files, GWAS summary statistics, causal-variant lists, variant-ID mappings, simulation configuration files, and available dataset-specific evaluation outputs.

Dataset-specific genetic and privacy-related evaluations are available in the Evaluation tab. These include chromosome-specific PCA and LD analyses for all cohort sizes, as well as Membership Inference Attack (MIA) and dataset-level kinship evaluations for the 10,000-sample datasets.

The Global evaluation tab provides an overview of genetic fidelity, causal-variant recovery, and computational scalability across populations and cohort sizes.

Full methodological details describing the generation and validation of these synthetic genotype-phenotype datasets will be provided in the accompanying publication.

The source code used to generate and evaluate the datasets described in this study is publicly available at https://github.com/EuracBiomedicalResearch/synthetic_data .


Phenotype mapping

Correspondence between trait number used in file names (for example phenotype 1) and the actual phenotype.