Human Genome Diversity Project
The Human Genome Diversity Project (HGDP) is an international research effort that systematically catalogs genetic variation across hundreds of global populations. Its core resource, the HGDP-CEPH panel, contains genome wide data from roughly 1,000 individuals representing 51 ethnolinguistically diverse groups. This guide is for bioinformaticians, population geneticists, evolutionary biologists, and clinical researchers who need a practical, source bounded framework for understanding, accessing, and responsibly analyzing HGDP data. NCBI Bookshelf offers foundational technical references on human genetic diversity and population genetics principles.
The HGDP began in the 1990s to address a critical gap: most genomic studies focused on European ancestry populations, leaving global variation underrepresented. By collecting samples from indigenous and remote populations, the project provides a unique window into human history, migration, and adaptation. Today its data underpin studies in disease genetics, admixture mapping, and natural selection. EMBL-EBI Training provides courses on population genetics and diversity analysis that complement HGDP use.
At a Glance
| Aspect | Description |
|---|---|
| Core Goal | Characterize human genetic diversity across global populations |
| Primary Dataset | HGDP-CEPH panel: ~1,000 samples from 51 populations |
| Key Methods | SNP genotyping, whole genome sequencing, population structure analysis |
| Major Applications | Admixture mapping, selection scans, rare variant interpretation, history inference |
| Common Pitfalls | Ignoring population stratification, underpowered per group sample sizes, batch effects |
| Access Points | NCBI Sequence Read Archive, CEPH Foundation, HGDP browser |
Core Concepts and Rationale
Human genetic diversity is the raw material of evolution and a critical confounder in medical genomics. The HGDP was designed to capture the full spectrum of variation that arose through migration, bottlenecks, admixture, and local adaptation. Unlike large cohort studies that oversample urban populations, the HGDP focuses on groups with distinct linguistic and geographic histories. PubMed source 7 demonstrates how ancient genomes from Neolithic contexts can be compared with modern HGDP data to trace ancestry and pathogen driven selection over millennia.
The project uses two main data types. First, high density SNP arrays (usually Illumina platforms) typed across the panel. Second, whole genome sequencing for a subset of individuals. These data allow researchers to compute allele frequencies, identify private variants, and infer population structure. Understanding the rationale for each data type helps you choose the right resource for your research question. The HGDP is not a clinical dataset, but it serves as a reference panel for imputation and admixture mapping in diverse cohorts. PubMed source 6 exemplifies how common and rare variant analyses across population scale biobanks leverage diversity references to uncover genetic architecture of complex diseases like lung cancer.
Decision Points for Using HGDP Data
Before integrating HGDP data into your analysis, evaluate several decision points. First, determine whether the HGDP panel matches the ancestry composition of your study population. If your cohort includes individuals from populations represented in the HGDP (such as Yoruba, Han Chinese, or Sardinian), the panel is a strong reference. If your population is admixed or missing from the panel, consider alternative references or methods for local ancestry inference. PubMed source 9 specifically addresses the challenges of local ancestry inference when reference panels are poorly matched, providing practical guidance on how to adjust.
Second, consider the variant frequency spectrum. HGDP SNP arrays tend to capture common variants (minor allele frequency > 5%) due to ascertainment bias from the chip design. For rare variant analysis, whole genome sequencing data from the HGDP or supplementary resources may be necessary. Third, account for sample size. Some HGDP populations have fewer than 20 individuals. This limits statistical power for population specific association tests. For such groups, you may pool closely related populations or treat them as meta populations.
Finally, decide whether to use raw HGDP genotype calls or precomputed principal components and ancestry estimates. Precomputed resources reduce computational burden but may not reflect your study's specific quality filters. Galaxy Training Network offers workflows for computing population structure from HGDP style data that you can adapt to your needs.
Practical Workflow for HGDP Analysis
A robust analysis workflow for HGDP data proceeds through six main stages.
Step 1: Obtain data. Download the HGDP data from the NCBI Sequence Read Archive for sequencing data or from the CEPH Foundation for genotyping call sets. Verify the version and release notes to ensure you are using the most current annotation.
Step 2: Preprocess and quality control. Convert files to standard formats (PLINK, VCF). Apply standard filters: call rate per sample and per marker (> 95%), Hardy Weinberg equilibrium (p > 1e 6), and minor allele frequency (> 1%). Remove samples with high heterozygosity or cryptic relatedness. Bioconductor provides R packages such as SNPedia and gdsfmt for these preprocessing steps.
Step 3: Merge with study data. If you are comparing HGDP with a target cohort, merge datasets on overlapping SNP sets. Perform strand alignment and remove ambiguous strand SNPs. Assess merge success by inspecting per sample missing rates after merging.
Step 4: Ancestry inference. Run principal component analysis (PCA) on merged data. Use the first 10 20 components to visualize population clusters. Identify outliers that may indicate sample swaps or mislabeled ancestry. Run ADMIXTURE or similar model based clustering for supervised ancestry proportions. Cross validate to choose the appropriate number of ancestral populations (K).
Step 5: Downstream analysis. Depending on your research question, you may test for association with phenotype (using linear mixed models to control for structure), detect signatures of selection (using Fst, iHS, or XP EHH), or perform admixture mapping. PubMed source 10 illustrates how whole genome doubling in cancer research benefits from reference panels that include diverse ancestry, though the specific method must be adjusted for HGDP scale data.
Step 6: Interpretation and reporting. Document all quality control thresholds, population assignments, and software versions. Report effect sizes with appropriate confidence intervals. Do not assume that HGDP represents all global diversity. Discuss the limitations of your reference panel.
Quality Checks and Validation
Quality checks must be applied at every workflow stage. For the raw HGDP data, verify sample call rate and heterozygosity. Populations with high inbreeding (such as isolated island groups) may show elevated homozygosity that mimics genotyping error. Check that SNP positions align to the correct reference genome build (GRCh37 or GRCh38). Use concordance checks with control samples if available.
For merged datasets, confirm that no systematic bias has been introduced by differences in genotyping platforms. PCA should show expected clustering, any large deviation may indicate batch effects. A useful validation step is to compute Fst between HGDP populations and your study populations. Expected Fst values for broad continental groups are well documented (e.g., ~0.15 between African and European). Check that your values match published benchmarks. PubMed source 8 provides an example of using population genetic metrics to infer purifying selection over long timescales, a method that relies on high quality diversity data.
Common Mistakes
One frequent error is treating HGDP populations as discrete, non admixed units. In reality, many HGDP individuals show evidence of recent admixture. Forcing a discrete ancestry model can produce artefactual results. Use continuous ancestry models (PCA) or multiple K values in ADMIXTURE.
Another mistake is using HGDP as a universal reference without considering the source of the data. The HGDP samples were collected in the 1990s, and some populations have changed dramatically due to migration. Do not assume that a sample labeled "Yoruba" reflects modern Yoruba allele frequencies perfectly. Always check for temporal and geographic consistency.
A third mistake is ignoring the ethical context. The HGDP has faced controversies regarding informed consent and data sharing. Use the data responsibly, respect the wishes of participating communities, and cite the project appropriately. Do not make claims about population susceptibility to disease without careful contextualization.
Finally, avoid over interpreting small population differences. With limited sample sizes per group, a subtle allele frequency difference may be due to sampling noise rather than biological distinction. Use bootstrapping or Bayesian methods to quantify uncertainty.
Limits and Uncertainty
The HGDP has several important limitations. Sample sizes per population range from 1 to 60 individuals, which reduces power for rare variant detection and for estimating population specific parameters like effective population size. The genotyping chips used have a strong ascertainment bias toward common variants discovered in European populations. This means many population specific or private variants are missed.
The panel also underrepresents many regions, such as South America, Oceania, and parts of Asia. It cannot substitute for more comprehensive efforts like the 1000 Genomes Project or the Human Genome Structural Variation Consortium. For disease mapping in clinical cohorts, the HGDP may not provide sufficient resolution for fine mapping causal variants.
When interpreting results, recognize that the HGDP is a snapshot of genetic variation from the late 20th century. It does not capture recent population dynamics, including urbanization, migration, and admixture that have reshaped many groups. Use the data as a reference, not as a definitive map of human diversity. PubMed source 11 shows that even highly resolved analyses like proteomics in cancer require integrating multiple data types, similarly, HGDP should be combined with other resources to draw robust conclusions.
Frequently Asked Questions
What is the difference between the HGDP and the 1000 Genomes Project? The HGDP focuses on indigenous and remote populations with deep cultural history, while the 1000 Genomes Project emphasizes broadly sampled global populations for variant cataloging. The HGDP has smaller sample sizes but greater ethnic diversity per population.
Can I use HGDP data for clinical risk prediction? No. The HGDP is a research reference, not a clinical dataset. It is not designed for individual risk assessment. Population specific effect sizes from HGDP may not translate to other cohorts without extensive validation.
How do I access the HGDP data? Data are publicly available from the NCBI Sequence Read Archive (for sequencing) and from the CEPH Foundation (for genotyping). You must agree to a data use agreement that respects the ethical guidelines of the original collection.
What are the ethical considerations of using HGDP? The project has been criticized for insufficient community engagement. Always cite the original collectors, use the data only for approved research, and avoid stigmatizing language about populations. Many resources like EMBL EBI training include ethics modules.
References and Further Reading
- NCBI Bookshelf General reference for human genetic diversity and population genetics concepts.
- EMBL-EBI Training Courses on population genetics, GWAS, and diversity analysis.
- Galaxy Training Network Workflows for ancestry inference and variant QC.
- Bioconductor Software packages for genomic data analysis in R.
- NCBI Sequence Read Archive Repository for HGDP sequencing data.
- Genetic architecture of lung cancer revealed by common and rare variant analyses across population scale biobanks Demonstrates use of diverse reference panels in disease genetics.
- Ancestry, admixture, and pathogens in contemporaneous Neolithic farmers and foragers Compares ancient and modern diversity data.
- Ten millennia of purifying selection on HLA B27 reveals an ancient epidemic scale burden of Spondyloarthritis Uses population genetic methods on long term evolution.
- Local ancestry inference with poorly matched reference panels Practical advice for handling mismatched references.
- Translating whole genome doubling into precision medicine in cancer Shows how genomic diversity factors into cancer research.