When a patient fills out a medical intake form, they are typically asked to check a box for race or ethnicity: White, Black, Asian, or Hispanic. In clinical practice, these discrete categories are frequently treated as biological switches. If a patient checks a certain box, a clinician might adjust a drug dosage, order a specific genetic screening, or calculate a baseline health risk.
However, human biology does not operate in discrete boxes. It operates on a continuous spectrum of genetic variation shaped by thousands of years of migration, isolation, and intermixing.
The core insight of a growing body of genomic research is that Self-Identified Race and Ethnicity (SIRE) is a socio-cultural construct that serves as a deeply flawed proxy for Genetically Inferred Ancestry (GIA). When precision medicine relies on categorical social labels instead of continuous biological metrics, it misclassifies genetic variants, misses critical disease carriers, and delivers inaccurate clinical predictions.
Why Categorical Proxies Fail
To understand why social categories fail as biological proxies, consider how the body processes medications. Drug metabolism is largely governed by specific genetic polymorphisms. If a population is highly admixed—meaning their ancestors came from geographically distinct groups—the alleles responsible for drug metabolism vary continuously alongside individual ancestry proportions, rather than aligning with external physical traits or social identities.
This breakdown is vividly apparent in pharmacogenomics. For example, the CYP3A5 gene encodes an enzyme critical for metabolizing more than half of all clinical drugs. In studies of the highly admixed Brazilian population, Suarez‐Kurtz et al. demonstrated that an individual's metabolic capacity is driven by their continuous genetic ancestry gradient, rendering self-reported skin color or geographic origin clinically useless for predicting drug responses (Pharmacogenomic Diversity among Brazilians: Influence of Ancestry, Self-Reported Color, and Geographical Origin; Global Pharmacogenomics: Distribution of CYP3A5 Polymorphisms and Phenotypes in the Brazilian Population). Individuals who identify with the same social or census category frequently possess vastly different drug-metabolizing profiles.
[Social Construct: SIRE] --> Discrete Boxes: [ White ] [ Black ] [ Hispanic ]
| | |
v v v
[Biological Reality: GIA] --> Continuous Gradient: |--|---|---|---|---|---|---|--|
When clinicians use broad categories to predict risk, they introduce dangerous blind spots. In a study of genetic markers across Mexican and Brazilian cohorts, Bonifaz-Peña et al. found significant genomic overlap between self-identified groups, meaning that group-based assumptions regularly miscalculate a patient's risk for adverse drug reactions (Exploring the Distribution of Genetic Markers of Pharmacogenomics Relevance in Brazilian and Mexican Populations).
This mismatch is not limited to pharmacogenomics; it fundamentally compromises diagnostic screening. Traditional medical guidelines often restrict certain genetic carrier screenings to specific ethnic groups deemed "high-risk." However, when Kaseniit et al. evaluated over 93,000 individuals undergoing expanded carrier screening, they discovered that ethnicity-based guidelines missed the vast majority of disease carriers (Genetic ancestry analysis on >93,000 individuals undergoing expanded carrier screening reveals limitations of ethnicity-based medical guidelines). Because a person's actual inherited disease risk is a product of biological lineage rather than cultural identity, relying on self-reported ethnicity systematically overlooks critical biological risk factors in multi-ancestral individuals.
The Mechanics of the Continuum
Why is the divergence between social identity and genetic reality so pronounced? The answer lies in the fine-scale population structures created by historical migrations.
When populations are aggregated into broad categories like "African American" or "Hispanic," the intricate genetic variations resulting from distinct histories of forced migration, localized isolation, and varying degrees of admixture are completely erased. As Mathias et al. demonstrated through whole-genome sequencing of African Diaspora populations across the Western Hemisphere, there is an unprecedented, fine-scale continuum of African, European, and Native American gene flow that cannot be compressed into a single monolithic category without losing clinical resolution (A continuum of admixture in the Western Hemisphere revealed by the African Diaspora genome).
In diverse urban environments, this genomic complexity intensifies. Analyzing a diverse patient population in Manhattan, Tayo et al. showed that broad social labels fail to capture the underlying genetic diversity of individuals living in admixed communities, making broad social categories highly inaccurate tools for personalized medicine (Genetic Background of Patients from a University Medical Center in Manhattan: Implications for Personalized Medicine).
To untangle these factors, large-scale biobanks are shifting toward frameworks that explicitly decouple social identity from genetic heritage. By pairing genomic data with electronic health records (EHRs) in the UCLA ATLAS biobank, Johnson et al. (Leveraging genomic diversity for discovery in an electronic health record linked biobank: the UCLA ATLAS Community Health Initiative) and Ding et al. (The UCLA ATLAS Community Health Initiative: promoting precision health research in a diverse biobank) demonstrated that GIA captures subcontinental biological diversity that SIRE completely misses. Crucially, separating these two variables allows researchers to isolate the independent contributions of biological risk factors from the social and environmental determinants of health that racial categories often track.
The Clinical and Algorithmic Path Forward
If the biological insufficiency of racial categories is so well-documented, why does clinical medicine continue to rely on them?
The primary bottleneck is institutional and operational. A survey of genetics professionals conducted by Popejoy et al. revealed a troubling lack of standardized definitions and protocols for collecting diversity measures in clinical settings (Clinical genetics lacks standard definitions and protocols for the collection and use of diversity measures). Because direct genetic ancestry is rarely calculated from patient DNA during routine clinical visits, professionals frequently substitute self-reported race or medical record data as a proxy for biological background—a practice that leads to high levels of uncertainty and the potential misclassification of genetic variants.
Furthermore, when clinical trials aggregate diverse cohorts into broad, monolithic racial categories, they mask critical internal variations in drug efficacy. In a meta-analysis of heart failure treatments, Li et al. found that while certain therapies showed disparate outcomes between broad Black and White cohorts, treating these groups as uniform populations oversimplifies the underlying genetic and socioeconomic diversity that actually drives the variations (Clinical outcomes of pharmacological therapies for heart failure in Black vs. White populations: a meta-analysis of randomized controlled trials of heart failure treatment).
Even automated identity verification systems struggle when they rely on rigid, binary genetic markers without accounting for complex biological realities. As Hu et al. noted, biological conditions like chimerism from bone marrow transplants can cause a patient to carry a donor's genotype, triggering false positives in automated tracking systems that lack nuanced clinical metadata (Genetic sex validation for sample tracking in next-generation sequencing clinical testing).
To build an equitable framework for precision medicine, researchers are developing methods that isolate true biological drivers from social proxies across three distinct fronts:
- Cellular Mapping: Global initiatives like the Human Cell Atlas are structuring their research around direct biological variables—such as single-cell genetics, environmental exposures, and life experiences—rather than social constructs, ensuring that cellular variation is mapped without reinforcing racial profiling (The commitment of the human cell atlas to humanity).
- Zero-Shot Machine Learning: In medical AI, Khara et al. developed a deep learning model for mammographic density prediction that achieves equitable "zero-shot" generalization across diverse racial subgroups, demonstrating that models trained to focus on localized physiological features can perform accurately on populations entirely absent from the training data (Generalisable deep learning method for mammographic density prediction across imaging techniques and self-reported race).
- Probabilistic Inferencing: Advanced computational frameworks, such as the Convolutional Embedded Networks (CEN) proposed by Karim et al., allow for population-scale bio-ancestry inferencing directly from genetic variants, explicitly mapping the probabilistic gap between overlapping genetic distributions and socially constructed categories (Convolutional Embedded Networks for Population Scale Clustering and Bio-Ancestry Inferencing).
Ultimately, as emphasized by Caliebe et al., distinguishing social constructs from biological ancestry in genetic epidemiology is not merely a matter of semantic precision; it is a scientific requirement (Including diverse and admixed populations in genetic epidemiology research). Moving away from categorical proxies and toward continuous models of genetic ancestry is the only way to ensure that precision medicine is both scientifically accurate and equitably applied.
Key Foundational Work
Genetic Background of Patients from a University Medical Center in Manhattan: Implications for Personalized Medicine (2011) — contributed 2 concept elements to this direction.
Pharmacogenomic Diversity among Brazilians: Influence of Ancestry, Self-Reported Color, and Geographical Origin (2012) — contributed 2 concept elements to this direction.