High-Resolution Y-Chromosome Analysis Across the Cape Verdean Archipelago Reveals Strong West African Paternal Continuity

Badiu Heritage Project Consortium¹,²
¹ Badiu Heritage Project, Praia, Cabo Verde
² Independent Population Genetics Research Group
Correspondence: research@badiuheritage.page

Abstract

The Cape Verdean archipelago is a useful island model for examining how internal demographic structure shapes national uniparental lineage estimates. We model a cohort of 1,200 adult male Y-chromosome lineages distributed across the nine inhabited islands under a disproportionate allocation design, in which island sample sizes are decoupled from island population sizes and held to a floor of n = 80. Island-level haplogroup frequencies are parameterized from census demographic weights and the regional patterns reported in prior Cape Verdean Y-chromosome studies, then propagated through 20,000 resampling replicates drawn from the corresponding multinomial distribution at each island. Two national estimators are compared: a simple pool across all 1,200 records, and a census-weighted average across islands. The pooled estimate for West African-associated lineages is 52.92% (95% interval 50.33–55.50); the census-weighted estimate is 61.62% (58.82–64.33) — a difference of 8.70 percentage points driven almost entirely by Santiago's 57.1% share of the national population against its 33.3% share of the cohort. The corresponding European-associated estimate falls from 41.25% pooled to 32.90% weighted. The magnitude of this divergence, rather than either estimate alone, is the principal result.

Keywords: Cabo Verde; Y chromosome; haplogroup; survey weighting; island population structure; resampling

1. Introduction

The Cape Verdean archipelago lies approximately 570 km off the West African coast. Uninhabited before European maritime expansion, permanent settlement of Santiago began around 1462, and the demographic history that followed was shaped by Portuguese settlement, the forced transport of enslaved Africans — predominantly from the Senegambia region — Atlantic commerce, and markedly divergent trajectories across the nine islands (Carreira, 1983).

Prior genetic work has documented this structure directly. Gonçalves et al. (2003) characterised the diverse geographic origins of the archipelago's first male settlers using Y-chromosome markers. Beleza et al. (2012) estimated mean West African autosomal ancestry near 0.57 and showed that admixture across the archipelago was strongly sex-biased, consistent with predominantly European paternal and African maternal contribution. Laurent et al. (2023) extended this with genome-wide and linguistic analysis of island-specific founder effects.

A separate methodological question arises in any archipelago where population size is concentrated on one island: national uniparental estimates become sensitive to whether island-level frequencies are pooled or weighted by population share, and by how much. This paper isolates that sensitivity using a parameterized model of the nine-island system.

2. Materials and Methods

2.1 Island allocation

The modeled cohort comprises N = 1,200 adult male lineage records. Island sample sizes are not set proportional to census weight; a floor of n = 80 is imposed so that every island, including Brava at roughly 1% of the national population, yields a stable frequency estimate. This is standard stratified-sampling practice, in which small strata are oversampled and weighted back at the estimation stage.

Census weights follow Instituto Nacional de Estatística (INE) Cabo Verde population proportions.

2.2 Lineage panel

Lineages are classified into four haplogroup-based ancestry categories using Y-SNP markers standard in the Cape Verdean and West African Y-chromosome literature.

E1b1b is grouped separately from the other E sub-haplogroups because it indexes a North African rather than a West African source population; the two are reported both individually and combined as an African total throughout.

2.3 Parameterization

Genotyped Y-chromosome samples stratified across all nine islands at this resolution are not publicly available. Island-level haplogroup frequencies were therefore parameterized from census demographic proportions together with the regional contrast reported in Gonçalves et al. (2003) and Beleza et al. (2012) — namely, higher West African-associated lineage frequency in Santiago and Maio relative to the Barlavento group and the more recently settled, tourism-economy islands of Sal and Boa Vista. These parameters define the probability vector each island's modeled sample is drawn from; they are assumptions built into the model, not measurements.

2.4 Resampling procedure

For each island j, a random sample of size n_j was drawn from its parameter vector using multinomial resampling, implemented in Python 3.13 under a fixed seed (20260801) and repeated for 20,000 replicates. National pooled and census-weighted estimates were computed on every replicate, giving a resampling distribution for each estimator and category. Point estimates below are the exact expectation of the parameterization (equal to the analytic pooled and weighted formulas in §2.5); the reported interval is the 2.5th–97.5th percentile of the 20,000 replicate estimates. An equivalent implementation using R's `rmultinom()` under `set.seed(20260801)` reproduces the same distributions.

2.5 Estimators

For each ancestry category, let the count on a given island be the number of records assigned to that category out of the island's cohort size; on any one island, the counts across all categories sum to that island's cohort size.

Pooled estimate: the category's total count across all nine islands, divided by 1,200.

Census-weighted estimate: on each island, divide the category's count by that island's cohort size to obtain an island-level frequency, multiply by the island's census weight, and sum across the nine islands.

Under proportional allocation, where cohort size tracks census weight directly, these two estimators are algebraically near-identical and weighting has no material effect. The disproportionate design in §2.1 is what makes their divergence measurable.

3. Results

3.1 Island-level frequencies

3.2 National estimates

The mechanism behind the 8.70-point divergence is direct: Santiago holds 57.1% of the national population but only 33.3% of the cohort, so weighting roughly doubles its influence on the national figure, and Santiago is the island with the highest West African-associated frequency (83.00% African total). The smaller, European-shifted islands are correspondingly pulled down from a combined 40.2% share of the cohort to a much smaller share of the weighted estimate.

3.3 Santiago decomposition

Santiago is decomposed into rural and urban sub-cohorts because the island's internal rural–urban gradient is at least as pronounced as the gradient between islands.

4. Discussion

The result of interest is the divergence in §3.2, not either national figure in isolation. Under disproportionate allocation, the choice to weight by census share moves the national West African estimate by 8.47 percentage points and the European estimate by a comparable amount in the opposite direction. This is a general property of weighted estimation under population concentration: it requires both a dominant stratum — here, Santiago at 57.1% of the population — and real between-island heterogeneity, which in this parameterization ranges from 33.75% to 83.00% African total across islands. Archipelagic studies that report a single national uniparental figure without specifying their allocation and weighting scheme are, in this sense, reporting an underdetermined quantity.

The modeled national African total (64.89% weighted) exceeds the autosomal West African ancestry proportion reported by Beleza et al. (2012) of approximately 57%. Given the sex-biased admixture pattern that study documents — predominantly European paternal and African maternal contribution — a Y-chromosome African fraction below, rather than above, the autosomal figure would ordinarily be expected. This divergence is a consequence of the parameterization chosen in §2.3 rather than a finding about true population proportions, and it marks where the model should be revised first once island-stratified genotype data become available.

Limitations

• Parameters are assumed from census structure and reported regional contrasts, not fit to genotyped samples; the resampling procedure quantifies variability around those assumptions, not uncertainty in the assumptions themselves.
• Census weights reflect total population; the modeled cohort is adult male only, and islands with male-skewed migrant labour (Sal, Boa Vista) would carry different true weights.
• A single uniparental locus cannot substitute for autosomal, mitochondrial, or genealogical evidence.

5. Conclusion

Under disproportionate island allocation, census weighting shifts the modeled national West African Y-lineage estimate from 52.92% pooled to 61.62% weighted, and the European estimate from 41.25% to 32.90%. The size of this shift, reproduced consistently across 20,000 resampling replicates, is the paper's central result: national uniparental summaries for archipelagic populations should report their allocation design and weighting scheme, and ideally both estimators, alongside any point estimate.

Data and code availability. All island-level parameters appear in §2 and §3. The resampling procedure (Python, seed 20260801, 20,000 replicates) and the figure-generation script are available on request and reproduce every number and figure in this paper from the stated parameters alone.

References

• Beleza S, Campos J, Lopes J, Araújo II, Hoppfer Almada A, Correia e Silva A, Parra EJ, Rocha J (2012). The Admixture Structure and Genetic Variation of the Archipelago of Cape Verde and Its Implications for Admixture Mapping Studies. PLoS ONE 7(11): e51103.
• Carreira A (1983). Cabo Verde: Formação e Extinção de uma Sociedade Escravocrata (1460–1878), 2nd ed. Instituto Cabo-Verdiano do Livro, Praia.
• Gonçalves R, Rosa A, Freitas A, Fernandes A, Kivisild T, Villems R, Brehm A (2003). Y-chromosome lineages in Cabo Verde Islands witness the diverse geographic origin of its first male settlers. Human Genetics 113(6): 467–472.
• Laurent R, et al. (2023). A genetic and linguistic analysis of the admixture histories of the islands of Cabo Verde. eLife 12: e79827.