Research article · Descriptive population genetics
Abstract
Background. Cabo Verde's islands have distinct settlement and migration histories. Unequal island representation can affect pooled Y-chromosome lineage frequencies.
Methods. Buccal-swab samples collected between March 2023 and February 2025 were analysed using Y-STR profiling and targeted Y-SNP testing. The analytical sample comprised 1,257 adult male participants grouped by their paternal grandfather's reported island of birth. We describe observed frequencies and present a census-resident-standardised summary of the combined A1a + E1a + E1b1a + E2b category using 2021 island resident-population shares. Equal-island weights and an illustrative weight set of undocumented origin were sensitivity analyses. Island intervals use Wilson scores; weighted estimates and paired differences use conditional fixed-stratum intervals.
Results. E1b1a was the largest category (465/1,257; 36.99%), followed by R1b (318/1,257; 25.30%) and E1b1b (184/1,257; 14.64%). The combined category accounted for 666/1,257 participants (52.98%). Its census-resident-standardised frequency was 63.17% (conditional 95% CI 60.93–65.40), exceeding the pooled frequency by 10.18 percentage points (paired 95% CI 9.01–11.35). Equal-island weighting gave 43.48%; the undocumented Table 1 scenario gave 63.89%. The latter exceeded the census-standardised result by 0.73 points (paired 95% CI 0.63–0.82).
Conclusions. Island weighting materially changes the combined summary. Census-resident standardisation provides a documented comparison distribution, not a validated estimate of national paternal-lineage frequencies: recruitment was non-probability-based, and resident population shares do not represent paternal-grandfather-origin shares.
Keywords: Cabo Verde; Y chromosome; haplogroup; island variation; population weighting
1. Introduction
Cabo Verde's history includes the forced movement of enslaved Africans, European colonisation, Atlantic commerce, and migration between islands. These processes did not affect every island in the same way. Genetic studies have consequently examined both the archipelago's shared history and differences among its island populations (Gonçalves et al., 2003; Beleza et al., 2012; Laurent et al., 2023).
For the Badiu Nation Research Project, this island-level perspective matters: a single archipelago-wide percentage can conceal Santiago's particular lineage composition. It is also important to distinguish the people represented in a dataset from the population to which an estimate is intended to apply.
Y-chromosome haplogroups describe branches within a hierarchical phylogeny of the non-recombining region of the Y chromosome. They provide a lineage-specific perspective on population history, distinct from genome-wide ancestry estimates (Y Chromosome Consortium, 2002; Beleza et al., 2012).
The objective of this study was to describe island-level variation in the supplied haplogroup counts and compare pooled frequencies with census-resident standardisation and alternative island-weighting schemes.
2. Materials and methods
2.1 Participants, sample collection, and island assignment
Samples were collected between March 2023 and February 2025 from adult male participants of Cape Verdean paternal ancestry. Participants were assigned to an island according to the reported birthplace of their paternal grandfather. Cases with uncertain paternal-island origin were excluded from island-specific analyses. Island categories therefore represent reported paternal geographic origin, rather than the participant's birthplace, current residence, or sample-collection location.
Buccal-swab samples were obtained using sterile collection kits and processed for Y-chromosome analysis as described below.
Known close paternal relatives and duplicate records were excluded during quality control. Additional exclusions covered incomplete ancestry information, failed genotyping, and other predefined quality-control criteria. The final analytical sample comprised 1,257 individuals across nine island categories. Each participant was included once and assigned to a single paternal-island category.
The frequency analysis used the resulting aggregate haplogroup counts. Island totals and category totals were checked for arithmetic consistency. Earlier publications provided comparative context rather than the source of these counts.
2.2 Y-chromosome genotyping and quality control
Y-STR haplotypes were generated with the Applied Biosystems Yfiler Plus PCR Amplification Kit (Thermo Fisher Scientific). The kit amplifies 27 Y-STR markers, including the two-copy systems DYS385a/b and DYF387S1a/b (Thermo Fisher Scientific, Yfiler Plus user guide). The marker set comprised DYS576, DYS389I, DYS635, DYS389II, DYS627, DYS460, DYS458, DYS19, Y-GATA-H4, DYS448, DYS391, DYS456, DYS390, DYS438, DYS392, DYS518, DYS570, DYS437, DYS385a/b, DYS449, DYS393, DYS439, DYS481, DYF387S1a/b, and DYS533.
Y-STR profiles were used for preliminary lineage assessment and quality control. Final haplogroup assignments were based on targeted Y-SNP genotyping. The hierarchical panel included M31 for A-lineage resolution; M33, M132, M2, M35, M215, M78, M81, M123, M75, M54, M85, and M98 for E-lineage resolution; M201, P15, and L30 for G; M170, P37.2, and M423 for I-lineage resolution; M304, M267, and M172 for J; M207, M343, and M269 for R-lineage resolution; and M70 for T. Downstream markers were tested after confirmation of the corresponding upstream branch. This list describes assay targets, not positive findings for every listed marker.
Targeted Y-SNP genotyping was performed with custom multiplex assays based on the Applied Biosystems SNaPshot Multiplex Kit (Thermo Fisher Scientific). SNP loci were arranged into hierarchical multiplexes according to their phylogenetic positions. Following locus-specific PCR amplification, alleles were determined by single-base primer extension. Extension products were separated by capillary electrophoresis on an Applied Biosystems 3500 Genetic Analyzer, and genotypes were reviewed using GeneMapper Software. Ambiguous calls were repeated before haplogroup assignment. The manufacturer's documentation describes the single-base-extension workflow, compatibility with 3500-series instruments, and GeneMapper analysis (Thermo Fisher Scientific, SNaPshot Multiplex Kit).
Haplogroups were assigned using the International Society of Genetic Genealogy Y-DNA Haplogroup Tree, version 15.73, dated 11 July 2020 (ISOGG, 2020). Classification proceeded from upstream defining SNPs to the most downstream informative marker tested in each sample. For population-level analysis, SNP-defined assignments were consolidated into the ten study reporting categories: A1a, E1a, E1b1a, E1b1b, E2b, R1b, I2a, G2a, J, and T. The defining SNP result and the study reporting label were retained together in the analysis file. Frequencies are reported at this grouped level rather than for every SNP-defined branch tested.
Profiles were accepted when they met predefined peak-height, allele-balance, and completeness criteria. Incomplete or ambiguous profiles were repeated, and specimens that continued to fail quality-control requirements were excluded. Positive and negative controls were included in each amplification batch. Concordance between the preliminary Y-STR lineage assessment and targeted Y-SNP results was reviewed before final assignment; where they differed, the SNP-defined lineage determined the final classification.
2.3 Pooled frequencies, census standardisation, and sensitivity analysis
Island proportions were calculated as pᵢ = xᵢ/nᵢ; the pooled proportion was Σxᵢ/Σnᵢ. Weighted summaries were calculated as p_w = Σwᵢpᵢ, with weights summing to one.
The main standardised summary for A1a + E1a + E1b1a + E2b uses 2021 island resident-population shares: wᵢ = Nᵢ/491,233. INE's definitive-results announcement, dated 4 April 2022, reports 505,044 people counted, including 491,233 residents; it does not list island totals. The nine integer island inputs were retrieved from the Pop_Res_2021 field, labelled “População Residente (2021)”, in INE's ILHAS layer (ID 1), within the enterprise-indicator service RE2023_INDICADORES_GERAIS_EMPRESAS_ILHA. Supplement S3 and the archived census-input file record the values, source, query, and retrieval date. Santa Luzia has a null population value in that layer and is outside the nine inhabited-island groups; it was not converted to zero.
These are all-resident, all-age, both-sex weights, not weights for adult men or paternal-grandfather origins. Applying them to paternal-origin strata defines a descriptive standardisation to a documented comparison distribution. It does not establish a nationally representative paternal-lineage estimate.
Sensitivity analyses use equal island weights and the Table 1 weights. The demographic source, year, and reference population of Table 1 remain undocumented; it is an illustrative scenario, not a validated population standard.
Table 1. Paternal-island sample sizes and illustrative scenario weights of undocumented origin.
| Island | Reported assignments | Scenario weight |
|---|---|---|
| Santiago | 417 | 0.571 |
| São Vicente | 199 | 0.156 |
| Santo Antão | 103 | 0.070 |
| Fogo | 107 | 0.064 |
| Sal | 111 | 0.064 |
| Boa Vista | 82 | 0.030 |
| São Nicolau | 89 | 0.022 |
| Maio | 76 | 0.013 |
| Brava | 73 | 0.010 |
| Total | 1,257 | 1.000 |
Island frequencies use 95% Wilson score intervals. Weighted intervals use the conditional fixed-stratum variance Σwᵢ²pᵢ(1 − pᵢ)/nᵢ and the estimate ± 1.959964 standard errors. This assumes independent Bernoulli sampling within paternal-island groups and known weights. Because recruitment was community-based rather than probability-based, these intervals describe variability under that model only; they do not capture recruitment bias, residual relatedness, uncertainty in weights, or resident-versus-paternal-origin differences. Excluding known close relatives does not establish independence.
For two weight sets applied to the same observations, paired differences use variance Σ(wᵢ − vᵢ)²pᵢ(1 − pᵢ)/nᵢ. The fixed-stratum pooled and pooled Wilson intervals represent different sampling models and are labelled separately. Their variance ratio is not an established survey design effect.
Combined-category heterogeneity was assessed using Pearson's χ² test on the 9 × 2 table and Cramér's V; all expected counts exceeded 5. No full 9 × 10 exact or Monte Carlo test is reported. Any future Monte Carlo test must specify its statistic, seed, draw count B, and estimator. Using (b + 1)/(B + 1), where b includes ties, 10,000 draws with zero exceedances give approximately 0.0001; they do not alone establish a true p-value below 0.0001.
All results are calculated from unrounded script output. Displayed percentages and differences use two decimals for auditability, not to imply that level of sampling precision.
2.4 Combined categories and geographic interpretation
Individual haplogroup frequencies were the primary outcomes. A1a + E1a + E1b1a + E2b was also evaluated as a combined analytical category. Additional grouped totals were reported by their constituent haplogroups. These combinations describe the distribution of assignments rather than continental ancestry proportions.
Geographic interpretation depends on branch resolution and comparative populations. For example, the identification of R-V88 in African populations demonstrates why broad R1b assignments cannot uniformly be equated with European ancestry (Cruciani et al., 2010). This example concerns interpretation of the classification, not an R-V88 finding in the present dataset.
3. Results
3.1 Reported island-level haplogroup counts
The dataset comprised 1,257 assignments, with paternal-island group sizes ranging from 73 for Brava to 417 for Santiago.
Table 2. Y-chromosome haplogroup counts by reported birthplace of the paternal grandfather.
| Island | n | A1a | E1a | E1b1a | E1b1b | E2b | R1b | I2a | G2a | J | T |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Santiago | 417 | 26 | 72 | 265 | 48 | 6 | 0 | 0 | 0 | 0 | 0 |
| São Vicente | 199 | 3 | 8 | 34 | 28 | 2 | 96 | 9 | 3 | 14 | 2 |
| Santo Antão | 103 | 1 | 4 | 27 | 23 | 0 | 39 | 1 | 1 | 5 | 2 |
| Fogo | 107 | 1 | 8 | 6 | 27 | 0 | 48 | 2 | 6 | 9 | 0 |
| Sal | 111 | 5 | 14 | 38 | 6 | 1 | 33 | 4 | 1 | 7 | 2 |
| Boa Vista | 82 | 0 | 1 | 28 | 19 | 0 | 27 | 1 | 3 | 3 | 0 |
| São Nicolau | 89 | 4 | 8 | 26 | 5 | 0 | 40 | 3 | 0 | 3 | 0 |
| Maio | 76 | 6 | 14 | 36 | 6 | 1 | 12 | 1 | 0 | 0 | 0 |
| Brava | 73 | 0 | 16 | 5 | 22 | 0 | 23 | 4 | 1 | 2 | 0 |
| Total | 1,257 | 46 | 145 | 465 | 184 | 10 | 318 | 25 | 15 | 43 | 6 |
3.2 Haplogroup composition
E1b1a is the largest pooled category at 36.99%. Under the specified weighting scenario, its frequency increases to 45.69%. R1b changes from 25.30% to 17.45%, while E1b1b changes from 14.64% to 13.51%.
Table 3. Pooled and scenario-weighted haplogroup frequencies.
| Reported haplogroup | Count | Pooled % | Scenario-weighted % |
|---|---|---|---|
| A1a | 46 | 3.66 | 4.41 |
| E1a | 145 | 11.54 | 12.74 |
| E1b1a | 465 | 36.99 | 45.69 |
| E1b1b | 184 | 14.64 | 13.51 |
| E2b | 10 | 0.80 | 1.05 |
| R1b | 318 | 25.30 | 17.45 |
| I2a | 25 | 1.99 | 1.31 |
| G2a | 15 | 1.19 | 0.84 |
| J | 43 | 3.42 | 2.59 |
| T | 6 | 0.48 | 0.41 |
Percentages may not sum to exactly 100% after rounding.
3.3 Island variation in the combined category
The combined A1a + E1a + E1b1a + E2b count is 666. Its frequency is highest on Santiago and lowest on Fogo. Heterogeneity across paternal-island groups was substantial (χ² = 410.98, 8 degrees of freedom; p = 8.37 × 10⁻⁸⁴; Cramér's V = 0.572). This measures association in the supplied counts under the independence assumption.
Table 4. Island frequencies of the combined A1a + E1a + E1b1a + E2b category.
| Island | Combined count | Island total | Combined % | 95% Wilson CI (%) |
|---|---|---|---|---|
| Santiago | 369 | 417 | 88.49 | 85.07–91.21 |
| São Vicente | 47 | 199 | 23.62 | 18.25–29.98 |
| Santo Antão | 32 | 103 | 31.07 | 22.95–40.55 |
| Fogo | 15 | 107 | 14.02 | 8.68–21.85 |
| Sal | 58 | 111 | 52.25 | 43.04–61.31 |
| Boa Vista | 29 | 82 | 35.37 | 25.89–46.16 |
| São Nicolau | 38 | 89 | 42.70 | 32.93–53.06 |
| Maio | 57 | 76 | 75.00 | 64.22–83.37 |
| Brava | 21 | 73 | 28.77 | 19.65–40.01 |
No R1b, I2a, G2a, J, or T assignments were observed among the 417 Santiago participants. For all five combined, the upper limit of the two-sided 95% exact interval is 0.88%; the one-sided 95% upper bound is 0.72%. These limits assume independent observations and do not establish population absence.
3.4 Effect of island weighting
Santiago contributes 33.17% of the assignments but receives 57.10% of the scenario weight. Because the combined category is especially frequent there, the weighted result is higher. This reflects changing island contributions, not discovering additional lineages or changing anyone's classification.
Table 5. Changes in grouped frequencies under the island-weighting scenario.
| Categories combined | Pooled count | Pooled % | Scenario-weighted % | Difference (percentage points) |
|---|---|---|---|---|
| A1a + E1a + E1b1a + E2b | 666 | 52.98 | 63.89 | +10.91 |
| E1b1b | 184 | 14.64 | 13.51 | −1.12 |
| R1b + I2a + G2a | 358 | 28.48 | 19.60 | −8.88 |
| J + T | 49 | 3.90 | 3.00 | −0.90 |
| All reported A and E categories (subtotal) | 850 | 67.62 | 77.41 | +9.78 |
The subtotal overlaps the first two rows and must not be added to the other rows. Differences are calculated from unrounded values, so subtracting displayed rounded percentages may give a slightly different result.
3.5 Conditional uncertainty and sensitivity to the weight set
Table 6. Combined category under four weighting schemes. Intervals are conditional fixed-stratum intervals.
| Weighting scheme | Combined % | 95% CI (%) | Difference vs pooled (pp) | Paired 95% CI for difference (pp) |
|---|---|---|---|---|
| Pooled (sample-size weights) | 52.98 | 50.72–55.25 | — | — |
| Table 1 scenario | 63.89 | 61.65–66.13 | +10.91 | 9.70–12.12 |
| 2021 census resident populations | 63.17 | 60.93–65.40 | +10.18 | 9.01–11.35 |
| Equal island weights | 43.48 | 40.60–46.35 | −9.51 | −10.78 to −8.24 |
The Table 1 scenario exceeds the census-weighted result by 0.73 percentage points (paired 95% CI 0.63–0.82), compared with its 10.91-point difference from the pooled result. This is a numerical sensitivity check; it does not establish the source of the Table 1 weights or validate either set for the paternal-origin population. Equal weighting answers a different question by assigning every island the same contribution.
The conditional variance ratio between the fixed-stratum pooled calculation and the simple pooled-binomial calculation is 0.673. This is a comparison of sampling models, not an established survey design effect or a genetic variance component.
4. Discussion
4.1 Island composition and comparison with previous studies
Census-resident standardisation increases the combined A1a, E1a, E1b1a, and E2b frequency from 52.98% to 63.17%, a difference of 10.18 percentage points. The illustrative Table 1 scenario gives 63.89%, while equal-island weighting gives 43.48%. These differences arise from the distribution used to combine the same island frequencies.
The distinction between pooled and weighted frequencies is compositional: the same island-level values are combined using different weights. Population inference additionally depends on suitable benchmarks and a defensible model of sample selection, as established in research on calibration of non-probability samples (Chen et al., 2018). Here, interpretation is restricted to the supplied aggregate dataset and specified weights.
Beleza et al. (2012) reported E1b1a at 18.8% and R1b1b2 at 42.7% among 431 men from six islands. The present table reports E1b1a at 36.99% and the broader R1b category at 25.30%. These are descriptive comparisons, not directly matched estimates: island coverage, allocation, and category resolution differ.
The difference already exists in the unweighted figures, so weighting cannot explain it on its own. The present study groups participants by paternal grandfather's birthplace. Comparisons with samples classified by participant birthplace or residence require harmonisation of the island-assignment rule as well as haplogroup definitions, recruitment, collection dates, and treatment of relatives. These are factors to investigate, not demonstrated causes of the difference.
The island-specific approach is also consistent with the distinct admixture histories examined by Laurent et al. (2023), although their genomic and linguistic analyses address different outcomes from the haplogroup frequencies reported here.
4.2 Lineages, history, and Badiu heritage
The non-recombining Y-chromosome phylogeny traces a direct paternal lineage (Y Chromosome Consortium, 2002). In Cabo Verde, analyses of Y-chromosomal, autosomal, and X-chromosomal markers have yielded distinct perspectives on population history and sex-biased admixture (Beleza et al., 2012). These different inheritance systems should therefore be interpreted as complementary, rather than interchangeable, descriptions of ancestry.
For Badiu heritage, the value here is the attention to Santiago and to variation concealed by a single archipelago-wide average. More detailed genetic classifications, family histories, and historical records could help connect particular lineages with that history. These aggregate counts do not yet establish those connections, and genetic categories do not determine who belongs to Badiu culture.
5. Conclusion
The dataset contains 1,257 reported Y-chromosome assignments, with E1b1a as its largest category. A1a + E1a + E1b1a + E2b accounts for 52.98% of the pooled data and 63.17% after standardisation to 2021 island resident-population shares. Santiago has the highest island-level frequency of this combined category, at 88.49%.
The census-standardised difference of 10.18 percentage points demonstrates sensitivity to island allocation. It does not correct recruitment bias or establish the distribution of paternal origins in the population. Table 1 remains an illustrative scenario of undocumented origin.
Data and reproducibility. Aggregate counts, category definitions, and scenario weights are provided above. Statistical supplement v1.0.0, dated 30 September 2026, contains the count CSV, frozen aggregate JSON, standalone calculation script, full-precision results, Tables S1–S7, and a SHA-256 checksum manifest. Its script is the numerical source for the added intervals and sensitivity results. The supplement and this article are archived on Zenodo as version 3.0 (doi:10.5281/zenodo.23484843), which replaces the version 2 PDF; doi:10.5281/zenodo.22745203 cites all versions and resolves to the latest. Data, tables, and documentation are licensed CC BY 4.0; the script is MIT-licensed. Node.js v24.15.0 was used for verification.
References
- Instituto Nacional de Estatística, Cabo Verde (2022). INE apresenta os resultados definitivos do V Recenseamento Geral da População e Habitação (RGPH-2021). 4 April 2022. Results announcement. The announcement supplies national totals, not island figures. Integer island resident counts were retrieved from Pop_Res_2021 in the ILHAS layer of INE's enterprise-indicator service.
- Chen JKT, Valliant RL, Elliott MR (2018). Model-assisted calibration of non-probability sample survey data using adaptive LASSO. Survey Methodology 44(1): 117–144. Full text.
- 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. Article.
- Cruciani F, Trombetta B, Sellitto D, et al. (2010). Human Y chromosome haplogroup R-V88: a paternal genetic record of early mid Holocene trans-Saharan connections and the spread of Chadic languages. European Journal of Human Genetics 18: 800–807. Article.
- 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. Article.
- Laurent R, Szpiech ZA, da Costa SS, et al. (2023). A genetic and linguistic analysis of the admixture histories of the islands of Cabo Verde. eLife 12: e79827. Article.
- Y Chromosome Consortium (2002). A nomenclature system for the tree of human Y-chromosomal binary haplogroups. Genome Research 12(2): 339–348. Article.
- Thermo Fisher Scientific. Yfiler Plus PCR Amplification Kit User Guide. Publication MAN0030230. Manufacturer documentation for kit composition. User guide.
- Thermo Fisher Scientific. SNaPshot Multiplex Kit. Catalogue 4323163. Manufacturer documentation for single-base-extension genotyping and instrument/software compatibility. Product documentation.
- International Society of Genetic Genealogy (ISOGG) (2020). Y-DNA Haplogroup Tree, version 15.73, 11 July 2020. Version history.