Hb F is the form of hemoglobin that dominates during fetal development. It consists of two alpha-globin chains and two gamma-globin chains, written as α2γ2. After birth, the body gradually replaces it with adult hemoglobin, α2β2, and in most healthy adults Hb F falls below 1 percent of total hemoglobin. In sickle cell disease, however, retaining more Hb F can be highly beneficial. The disease is caused by a mutation in the HBB gene that replaces the amino acid glutamic acid with valine in beta-globin, producing hemoglobin S. When oxygen levels fall, hemoglobin S molecules can stick together and form rigid polymers, distorting red blood cells into the characteristic sickle shape. Hb F interferes with this polymerization because gamma-globin-containing molecules do not participate in the same destructive assembly. Higher Hb F can therefore reduce red-cell rupture, blood-vessel blockage, severe pain and cumulative organ damage.

The review found that the strongest and most consistently replicated signals cluster in three genomic regions: BCL11A on chromosome 2, the intergenic region between HBS1L and MYB on chromosome 6, and the beta-globin gene cluster on chromosome 11, which includes the gamma-globin genes HBG1 and HBG2. BCL11A emerged as the most extensively studied locus, with 13 variants reported across 37 studies. Several of its variants, including rs1427407, rs4671393 and rs11886868, are located in an erythroid-specific enhancer, a regulatory DNA segment that controls gene activity in developing red blood cells. Hb F-increasing versions of these variants appear to weaken the enhancer’s ability to activate BCL11A. That matters because BCL11A normally acts as a powerful repressor of fetal hemoglobin after birth. When BCL11A levels decline, repression at the HBG1 and HBG2 promoters is eased, allowing gamma-globin production to persist.

The quantitative results reinforced the central role of the BCL11A enhancer, although they also exposed the difficulty of combining data from genetically diverse populations. Across nine studies, rs1427407 was associated with a pooled mean Hb F difference of 1.1, with a 95 percent confidence interval ranging from 0.287 to 1.914. The comparable pooled effects were 1.13 for rs4671393 and 1.229 for rs11886868, and all three reached statistical significance. Yet heterogeneity was very high, with I² values of 90.7, 95.2 and 87.9 percent, respectively. In meta-analysis, I² estimates the proportion of variation between studies that cannot be explained by sampling error alone. Such high values suggest that ancestry, local haplotypes, clinical characteristics, treatment exposure or differences in how Hb F was measured may substantially alter the effect of a variant. Some BCL11A variants even showed opposite associations in different populations, underscoring why a DNA change cannot always be interpreted in isolation from its surrounding genetic background.

The second major region, HBS1L-MYB, contained the largest number of reported variants: 28 across 25 studies. Many Hb F-increasing variants in this region appear to reduce expression of MYB, a regulator of red-cell development. Lower MYB activity can lengthen the cell cycle of erythroid progenitors, the immature cells that eventually become red blood cells. This extended developmental window may give gamma-globin expression more time to remain active, increasing the proportion of cells that contain Hb F. The variant rs4895441 produced one of the most consistent signals in the review, with a pooled mean difference of 0.356 and a 95 percent confidence interval of 0.181 to 0.532. Its I² value was 49.4 percent, considerably lower than for most of the other variants analyzed. By contrast, rs9399137 showed a directionally positive but statistically uncertain result, while rs28384513 produced a small suppressive effect that was remarkably consistent across five studies, with I² equal to zero.

The beta-globin cluster produced a particularly dramatic example of ancestry-dependent genetics. The HBG2 variant rs7482144, commonly known as the XmnI polymorphism, lies about 158 DNA letters upstream of the HBG2 gene promoter. Across 13 studies, it was associated with a pooled Hb F increase of 2.855, but the confidence interval was broad, from 0.626 to 5.083, and heterogeneity reached 96 percent. The variant’s effect depends heavily on the larger beta-globin haplotype in which it occurs. It is associated with especially high Hb F in the Arab-Indian haplotype, while its influence is weaker or nearly absent in some African populations carrying Bantu haplotypes. This genetic context helps explain why some individuals with sickle cell disease in parts of India and the Arabian Peninsula can have Hb F levels above 20 or even 30 percent, whereas many people of West African ancestry have much lower baseline levels. The difference is not simply a matter of one mutation, but of inherited regulatory combinations accumulated across the globin cluster.

The analysis also identified evidence for a wider network of biological influences beyond the three canonical regions. The most prominent was HMOX1, which encodes heme oxygenase-1, an enzyme involved in breaking down heme and responding to oxidative stress. The HMOX1 variant rs2071746 produced the largest pooled estimate among the variants included in the meta-analysis, with a mean difference of 4.453, but its I² value of 97.8 percent makes that number difficult to generalize. One possible mechanism is that increased heme oxygenase-1 activity generates carbon monoxide, which can activate soluble guanylate cyclase and raise cyclic GMP levels, a signaling pathway known to stimulate gamma-globin transcription. Other emerging genes included SIN3A, ZBTB7A, FOXO3, ANTXR1, BACH2 and HIF-1α. These genes touch several layers of regulation, including chromatin remodeling, transcriptional repression, oxygen sensing and erythroid stress responses. Most of these associations appeared in only one or two studies, so they are promising clues rather than established clinical markers.

The therapeutic implications are substantial because the genetic findings converge with modern efforts to reactivate fetal hemoglobin directly. BCL11A enhancer disruption is already being tested as a treatment strategy using CRISPR-Cas9 gene editing. By editing regulatory DNA in a patient’s blood-forming stem cells, researchers aim to reduce BCL11A activity specifically in red-cell precursors, releasing the brake on gamma-globin production. The review’s natural genetic associations provide an important biological validation of that approach: variants that weaken the same enhancer are linked to higher Hb F in people who have not undergone gene editing. The results could also support genotype-guided clinical trials. Patients carrying Hb F-promoting alleles might respond differently to hydroxyurea or other Hb F-inducing medicines than those carrying suppressive variants. However, the authors caution that no genetic test can yet replace clinical assessment. The evidence is uneven, effect sizes vary between populations, and many studies did not report results in a standardized form.

The review ultimately presents Hb F regulation as a multi-layered genetic network shaped by ancestry, haplotypes and interactions among regulatory proteins. It also highlights a serious equity problem: sub-Saharan Africa accounts for roughly 80 percent of the global sickle cell disease burden, yet African populations remain underrepresented in genomic research compared with African American, European and some Asian cohorts. The study’s authors argue that future work should prioritize large African genome-wide association studies, trans-ethnic fine-mapping and multi-omics analyses that connect DNA variants to chromatin accessibility and gene activity in erythroid cells. The current results are powerful because seven of the eight meta-analyzed variants showed directionally consistent associations, but they are not a universal genetic formula for sickle cell disease. Instead, they offer a framework for understanding why Hb F varies so widely between individuals—and a roadmap for developing treatments that could reproduce nature’s most effective protection against sickling.

The study focuses on admixture mapping, a statistical approach developed for populations whose genomes contain ancestry segments inherited from multiple continental or regional source populations. In an admixed individual, nearby stretches of DNA can have different ancestral origins. For example, one chromosome segment may be more closely related to West African reference populations, while another may reflect European, Indigenous American or East Asian ancestry. By estimating the ancestry of these local genomic segments and testing whether they are associated with a trait, researchers can search for genetic regions that help explain phenotypic variation. This differs from conventional genome-wide association studies, which generally test individual genetic variants across a population.

The distinction is important because a person’s overall or “global” ancestry may conceal the influence of particular genomic regions. Two people can have broadly similar ancestry proportions but carry different local ancestry patterns across their chromosomes. Admixture mapping takes advantage of this mosaic structure. If a trait is consistently associated with a particular ancestry at one genomic location, that signal may point to genetic variants, regulatory elements or biological pathways that influence the phenotype. However, local ancestry is not itself a biological cause. It is a marker for inherited genetic variation, and its interpretation requires careful attention to environmental exposures, socioeconomic conditions, access to health care and the historical forces that shaped population structure.

The All of Us Research Program provides an unusually broad setting for this kind of analysis. The program was created to assemble health, genomic, demographic, lifestyle and electronic medical-record data from a highly diverse group of participants in the United States. Rather than relying primarily on cohorts recruited for a single disease, All of Us is designed to support research across many conditions and traits. Its participants contribute information through surveys, clinical records, physical measurements and, for many, genomic testing. This combination allows investigators to study ancestry-related patterns alongside the social and medical context in which those patterns occur.

Mandla and colleagues use this resource to improve the resolution of comparisons between populations with different ancestral histories. Such comparisons have often been limited by the underrepresentation of non-European groups in genetic studies, uneven sample sizes and analytical methods that treat ancestry as a single genome-wide percentage. Those limitations can produce unstable associations or make biological differences appear larger or smaller than they really are. By incorporating local ancestry and a much larger, more heterogeneous dataset, the researchers can test whether observed phenotypic differences are linked to particular genomic regions rather than to broad labels assigned to entire populations.

The technical challenge is substantial. Before an admixture-mapping analysis can begin, scientists must infer local ancestry along each participant’s chromosomes. This typically involves comparing genetic markers with reference panels representing ancestral source populations and calculating the most likely ancestry state for successive genomic segments. The resulting data are then analyzed alongside traits such as disease risk, physiological measurements or other characteristics recorded in health records. Statistical models must account for relatedness among participants, sex, age, study site, global ancestry and other potential confounders. The goal is to identify ancestry-associated signals that remain meaningful after these factors are considered.

A major contribution of the work is its emphasis on improving the characterization of cross-population phenotypic differences rather than simply ranking groups by genetic risk. Population-level differences in a trait can arise through multiple pathways. They may reflect genetic variation, differences in diet or occupational exposure, unequal treatment within health systems, environmental pollution, stress, income, neighborhood conditions or historical patterns of migration. In many cases, these factors are correlated with ancestry, making them difficult to disentangle. Admixture mapping cannot solve that problem alone, but it can help identify genomic regions that warrant functional investigation while reducing the temptation to interpret every population difference as either purely genetic or purely social.

The findings also carry implications for precision medicine. Many genetic prediction models perform best in populations whose genomic data were used to build them, but less accurately in groups that have been historically excluded from research. This imbalance can affect disease-risk estimation, drug-response prediction and the interpretation of clinical genetic tests. A more detailed understanding of local ancestry may help researchers determine whether a genetic association operates similarly across populations or whether its effect changes depending on the surrounding genomic background. It may also reveal variants that were missed in earlier studies because they are uncommon in European-ancestry datasets but more frequent in other populations.

At the same time, the study underscores why ancestry-aware genomics must be communicated carefully. Genetic ancestry is not equivalent to race, ethnicity, nationality or culture, and no population is genetically uniform. Admixed genomes are shaped by many generations of movement and reproduction, while racial categories are social classifications that vary across time and place. The researchers’ approach is therefore most useful when it points toward specific biological mechanisms and is interpreted alongside detailed information about participants’ lived environments. Used responsibly, large-scale admixture mapping can expand the reach of genomic medicine without turning population labels into biological destiny.

The broader message from the All of Us analysis is that diversity is not merely a matter of representation or fairness, although both remain essential. It is also a scientific resource that can expose genetic signals, environmental interactions and disease mechanisms that would remain invisible in narrower datasets. By combining local ancestry inference with extensive health information, the study provides a framework for examining how inherited variation and social context intersect in real populations. The result is a more precise, more cautious and potentially more clinically useful picture of human phenotypic diversity—one that replaces simplistic population comparisons with a detailed view of the genomic mosaic carried by each individual.

Subject of Research: Large-scale admixture mapping, local genetic ancestry, and cross-population phenotypic differences using data from the All of Us Research Program.

Article Title: Large-scale admixture mapping in the All of Us Research Program improves the characterization of cross-population phenotypic differences.

Article References: Mandla, R., Shi, Z., Hou, K. et al. “Large-scale admixture mapping in the All of Us Research Program improves the characterization of cross-population phenotypic differences.” Nature Communications (2026). https://doi.org/10.1038/s41467-026-75515-6

Image Credits: AI Generated

DOI: 10.1038/s41467-026-75515-6

Keywords: admixture mapping, local ancestry, genetic ancestry, population genomics, All of Us Research Program, phenotypic diversity, precision medicine, human genetics, genomic medicine, cross-population differences