Geographic Sampling Bias in Genomic Surveillance

85 concepts68 papers1 proposal

When we look at a global map of viral mutations, a phylogenetic tree of a bacterial pathogen, or an environmental catalog of soil microbes, we assume we are looking at nature. In reality, we are often looking at a map of global sequencing infrastructure.

Genomic surveillance relies on the extraction, sequencing, and public sharing of genetic data. However, because sequencing capacity is deeply tied to regional wealth and institutional infrastructure, data collection is heavily skewed toward high-income nations and centralized clinical hubs. This phenomenon—geographic sampling bias—does not merely leave blank spaces on the map; it actively distorts our understanding of evolutionary biology, misleads predictive machine learning models, and compromises global public health interventions.

To understand why this bias occurs and how it warps our scientific consensus, we must look at how genomic data is processed and interpreted.


The Optical Illusion of Evolutionary Hubs

In phylogenetic analysis, the evolutionary distance between organisms is inferred by comparing genetic sequences and reconstructing the most likely sequence of mutations over time. When one region sequences thousands of viral samples per week while another sequences only a dozen, the resulting phylogenetic trees suffer from a severe mathematical artifact.

The region with dense sampling will naturally reveal a highly resolved, continuous web of minor genetic variations. In contrast, the under-sampled region will appear to have low viral diversity and slow evolutionary rates simply because intermediate mutations are never captured. As Zhengqiao Zhao et al. note in their work on SARS-CoV-2 classification in Genetic grouping of SARS-CoV-2 coronavirus sequences using informative subtype markers for pandemic spread visualization, varying regional sequencing capacities directly skew the perceived prevalence and evolutionary speed of viral lineages.

This sampling imbalance creates an optical illusion: well-monitored regions appear to be "hubs" of viral innovation and transmission, while under-monitored regions are treated as passive recipients. When genomic analysis is restricted to a single geographic region or heavily skewed toward a few hubs, the identified evolutionary patterns reflect local outbreaks rather than worldwide trends.


Creating Artificial Genetic Breaks

In ecology and biogeography, researchers use genetic data to determine whether a species exists as a single, well-connected population or if it has split into distinct sub-species due to physical barriers (like mountains or oceans).

When geographic sampling is sparse or uneven, continuous genetic variation across a landscape is easily misinterpreted. Imagine a species whose genetic traits change gradually from West to East—a phenomenon known as "Isolation by Distance." If scientists only collect samples from the far West and the far East, skipping the intermediate zones due to logistical or infrastructural constraints, the resulting data will show a massive genetic gap.

Actual Population (Continuous Gradient):
[A1] --- [A2] --- [A3] --- [A4] --- [A5] --- [A6]

Biased Sampling (Missing Intermediate Data):
[A1]                                         [A6]
  \___________________________________________/
            Artificial "Genetic Break"

In Bats of a feather: Range characteristics and wing morphology predict phylogeographic breaks in volant vertebrates, Kaiya L Provost et al. demonstrate how this exact artifact confounds ecological models. When aggregated databases contain geographic gaps, continuous genetic clines are frequently misidentified as discrete population breaks or speciation events. This makes it incredibly difficult for biologists to distinguish between true evolutionary barriers and mere artifacts of incomplete data collection.


The Failure of Western-Centric Medical Models

The consequences of this bias extend far beyond academic taxonomy; they directly impact drug discovery and vaccine design. Because public genomic databases are heavily dominated by sequences from Western nations, the medical models derived from them are fundamentally non-representative of the global population.

Consider the evolution of drug resistance in HIV-1. In Comparing mutational pathways to lopinavir resistance in HIV-1 subtypes B versus C, Susana Posada-Céspedes et al. highlight that the vast majority of viral genomic research has historically focused on subtype B, the strain most prevalent in Western nations. However, subtype B represents only a small fraction of global infections, while subtype C is far more widespread globally. The authors demonstrate that subtypes B and C follow entirely different mutational pathways to resist antiretroviral drugs. Consequently, diagnostic tools and clinical models optimized for Western data fail when applied to the global majority of patients.

Similarly, when designing diagnostic markers or identifying therapeutic targets for bacterial pathogens, a lack of balanced global sampling can lead scientists down the wrong path. In a comparative genomics study of Listeria monocytogenes, Bo Zhang et al. argue in Comparative genomics analysis to explore the biodiversity and mining novel target genes of Listeria monocytogenes strains from different regions that an overrepresentation of data from specific continents limits the precision of global evolutionary maps. Without representative data from all geographic regions, the "conserved" genes identified as targets for therapies or vaccines may fail to account for critical regional genetic variations.


Blinding Predictive Machine Learning Models

As computational biology moves toward predictive modeling—such as using machine learning to forecast which viral variants will emerge next—the underlying data quality becomes paramount. Machine learning models are notorious for treating correlation as causation; if they are trained on biased data, they will learn the bias rather than the biology.

In Dynamics-Based Feature Augmentation of Graph Neural Networks for Variant Emergence Prediction, Majd Al Aawar et al. address the challenge of predicting exactly when and where a new biological variant will arrive. They note that forecasting models rely entirely on high-quality genomic sequencing, which is non-uniformly distributed across the globe. When a Graph Neural Network (GNN) attempts to map the global spread of a pathogen, discrepancies in regional sequencing frequency create massive data blind spots. The model may mistake a lack of reporting for a lack of viral presence, completely compromising the reliability of early-warning public health systems in under-monitored regions.


Expanding the Global Genomic Horizon

The distortion caused by geographic sampling bias affects every level of genomic research, from large-scale environmental catalogs to human health. For instance, Antonio Pedro Camargo et al., in their analysis of the A global soil plasmidome resource unveils functional and ecological roles of plasmids in soil microbiomes, point out that even massive environmental genomic catalogs remain heavily skewed toward North America, hindering the discovery of unique genetic diversity in the Global South. When researchers actively work to fill these gaps—as seen in the gut microbiome research by Quanbin Dong et al. in Expanded gut microbial genomes from Chinese populations reveal population-specific genomic features related to human physiological traits—they routinely uncover thousands of entirely new species and deep genetic variations that were completely invisible to Western-centric databases.

To build genomic models that truly work globally, the scientific community must move past opportunistic sampling—simply analyzing whatever data happens to be available in public repositories—and transition toward systematic, decentralized, and globally distributed surveillance infrastructure. Only then can we ensure that our evolutionary maps reflect the true diversity of nature rather than the boundaries of global infrastructure.

Key Foundational Work

Two new species of the genus <i>Teuchophorus</i> (Diptera: Dolichopodidae) from Taiwan (2006) — contributed 2 concept elements to this direction.
History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 69 papers
  2. Version 1 · 2026-10-04 · 78 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_geographic_sampling_bias_genomic_surveil_7247fec1,
  title = {geographic sampling bias (genomic surveillance disparity)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/geographic-sampling-bias-genomic-surveillance-disparity/rdr_7247fec1f6e0477e9d07d4ea35c0cfa6},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 2, frozen text):

@misc{lacuna_geographic_sampling_bias_genomic_surveil_7247fec1_v2,
  title = {geographic sampling bias (genomic surveillance disparity)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/geographic-sampling-bias-genomic-surveillance-disparity/rdr_7247fec1f6e0477e9d07d4ea35c0cfa6/v/2},
  note = {Version 2, frozen 2026-10-04. Accessed 2026-11-21},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake