The missing diversity in human epigenomic studies

Abstract

Recent work has highlighted a lack of diversity in genomic studies. However, less attention has been given to epigenomics. Here, we show that epigenomic studies are lacking in diversity and propose several solutions to address this problem.

Research in diverse populations is critical for understanding disease etiology and risk. Several recent publications have highlighted the lack of racial or ethnic diversity in genetic studies and have called for more research in diverse populations^1,2. However, less attention has been given to epigenomics. Over the past ten years, great progress has been made in the understanding of regulatory elements through the efforts of the International Human Epigenome Consortium (IHEC), which mapped regulatory elements in a wide range of tissues and cell types, and made many of these datasets freely available to the scientific community³. This comprehensive catalogue of cis-regulatory elements and chromatin datasets has proved useful for different areas such as genomic variant annotation⁴, fine-mapping of genetic loci⁵, genome editing approaches⁵, and design of pipelines for single cell-sequencing analyses⁶.

Current information regarding the race or ethnicity of IHEC samples is sparse. We queried publicly available IHEC datasets for different statistical metrics relating to race/ethnicity and country of origin, finding only 42.7% of experiments reporting any race or ethnicity information (Supplementary Table 1, downloaded from https://www.encodeproject.org/; we used US-based ENCODE data as it was the only publicly available dataset within IHEC). Of the 5,048 publicly available experiments with race or ethnicity information, 87.1% (n=4,397) were labelled as “European”, 9.3% (n=470) were reported as African, African American or Black, 1.7% (n=87) were of Asian ancestry, and the remainder (1.9%, n=94) were of other ancestries or a combination of racial/ethnic identities, showing considerable disparity in the samples utilized for analysis (Supplementary Table 1). From 2009 to 2021, the cumulative number of experiments on “European” samples increased, far outpacing experiments on samples from other races and ethnicities (Figure 1). Although a set of experiments based on specific African populations (e.g. Luhya, Maasai, Mende, Esan, and Gambia) was posted in 2021, increasing the diversity of data available, populations from other geographic regions (e.g., South Asia, Middle East) remain underrepresented.

Figure 1: Diversity in epigenetic data samples over time:

Shown is the cumulative number of samples per year in publicly-available IHEC data (2009–2020, left panel). Different segments of the chart are color-coded by ethnicity as found in IHEC. Ethnicity information was obtained from the relevant studies. Given the large proportion of samples in individuals of European ethnicity as found in IHEC (red), a zoom-in chart is provided (right panel) showing the different cumulative sample numbers in non-European populations across the same timeframe.

The breadth of epigenomic assays and tissues used is substantially more extensive for Europeans than for other races/ethnicities. Among assays, ATAC-seq, DNase-seq, ChIP-seq and DNA methylation arrays show the highest degree of diversity with data from more than 6 populations (Supplementary Table 1, Figure 2). Although Hispanics were represented in relatively few experiments (n=60), a more comprehensive set of annotations across main assay types, such as RNA-seq, DNase-seq and ChIP-seq (including ChIP-seq for CTCF and histone H3 modifications) is available for them compared to other non-European populations. We also noted that data from non-European populations largely come from cell lines. Although valuable, the immortalization and serial passage of cell lines can lead to epigenetic changes that are not present in the primary cells and tissues⁷. The experiments conducted in primary tissues are overwhelmingly from “Europeans”, with few primary tissue experiments in non-Europeans. Given limited non-European primary tissue samples, any differences in tissue-specific regulatory elements across populations will be hard to evaluate.

Figure 2: Diversity in epigenetic data samples by assay:

the doughnut chart (lower panel) shows the total number of samples by assay (outer ring) and by ethnicity (inner ring). Given the large proportion of samples in individuals of European ethnicity as found in IHEC, a zoom-in area chart is provided (top panel) showing the different sample number by assay in non-European populations.

An essential question in characterizing regulatory elements across populations is the role of DNA sequence variants. The extent to which ancestry-related DNA sequence variants affect epigenetic modifications is unknown. However, there is evidence for widespread epigenetic variation between populations, particularly with regards to DNA methylation^8,9,10. While some sections of the epigenome are influenced by environmental exposures^11,12,13, many epigenetic changes are driven by changes in the DNA sequence^10,14,15,16. For example, twin studies have shown that the mean genetic heritability of DNA methylation is 19%, with some regions showing a heritability of over 90%¹⁵, suggesting that DNA methylation, particularly in those regions, is likely to be determined in large part by underlying genetic variants. Other studies have previously reported associations between individual ancestry-specific DNA sequence variants and DNA methylation differences between populations^17,18. Given this evidence, we anticipate that more associations between genotype, DNA methylation and ancestry may be uncovered in the future, which could potentially help explain population disparities in disease risk. In short, the role of ancestry-related DNA sequence variants in driving epigenetic variation needs to be explored further, especially in regard to disease-associated regions.

Epigenomic resources in diverse populations could contribute to annotating and interpreting disease-associated genomic regions. Genome-wide association studies (GWAS) have identified thousands of loci for various diseases and traits^19,20,21. However, many of these variants are located in non-coding regions of the genome with unclear functional consequences^4,22. Mapping these variants to the regulatory elements, including promoters, enhancers, and repressors, through epigenomic markers can provide important insights into possible functional mechanisms across a variety of tissues and cell types^4,23. The extent to which current epigenomic mapping resources, which are mostly European-centric, facilitate interpretation of GWAS loci in diverse populations is unknown. However, expanded epigenomic mapping data in diverse populations may improve the interpretation of disease-associated loci across populations^9,24 and offer additional insights. Expanded population-specific epigenomic maps may be particularly useful for annotating and fine-mapping variants in diseases with a higher burden in non-European populations, such as prostate cancer²⁵, hypertension²⁶, and chronic kidney disease²⁷.

In conclusion, additional research is warranted to evaluate the diversity in the epigenome across populations and determine the extent of population variability. Current efforts to increase representation in genomic research in diverse populations should be paired with similar efforts in epigenomics, which have, thus far, received less attention and scientific scrutiny. The posting of ancestry information, which could be inferred from sequencing or genotype array data, with existing epigenomic data could be beneficial in helping researchers understand the potential limitations for annotating and interpreting GWAS loci from different populations. Regarding IHEC, we recommend that participating consortia post genetic ancestry assignment inferred using reference genomes. While consortia may include self-reported race/ethnicity (for example in the US-based consortium reported here), we recommend analyses at the international scale first focus on genetic ancestry given the substantial challenges in standardizing race/ethnicity reporting across different countries. In addition, efforts to diversify IHEC participating countries should be promoted. Future studies should concentrate on generating high-quality data across diverse populations, using ancestry-specific reference genomes for aligning or mapping chromatin peaks from diverse populations, and developing DNA methylation arrays that adequately capture epigenomic diversity across populations. Improvement of the diversity of epigenomic resources will likely accelerate research addressing disease risk and health disparities across populations.