Galicia y la inmigración: ¿es la comunidad con más extranjeros?

Si es que todo tiene su explicación.

Whole-genome sequencing in Galicia reveals male-biased pre-Sarracinic North African ancestry, subtle population structure, and micro-geographic patterns of disease risk​

Galicia, located at the westernmost edge of Europe, has been reported to exhibit distinctive genetic traits compared to other Iberian populations. We present the first whole-genome sequencing (WGS) study of a Galician population (GALOMICS [GAL]; n = 91, 17.2 million variants; ), assessing its genetic variability in comparison with WGS data from other Spanish regions and continental populations (n = 1,078). Contrary to recent claims of extreme genetic stratification, the population structure of Galicia aligns with broader Iberian patterns, with one dominant cluster homogeneously distributed and four minor, more localized clusters. Genome-wide analyses of the Spanish National DNA Bank dataset (NDNAB; n = 453) support these findings, identifying only three Galician clusters, again with one overwhelmingly predominant. Phylogenetic analyses challenge earlier interpretations that placed Galicians at the deepest Iberian genetic node; instead, Galician clusters form terminal branches, suggesting recent diversification. Analysis of runs of homozygosity indicates slightly higher inbreeding compared to other European populations, primarily driven by the ‘Porto do Son’ cluster, which raises the regional average. We identified a significant North African/Middle Eastern autosomal ancestry component (13.5%–16.5%), despite its distance from historically Arab-influenced regions. Genomic analyses point to an admixture event ca. 620–670 CE that introduced North African/Middle Eastern ancestry into a largely European gene pool. The signal, likely stemming from trans-Mediterranean contacts predating the 711 CE Sarracinic incursion and well before the Reconquista, shows a subtle South-to-North decline, suggesting a southern entry route. Y-chromosome (21.2%) and mitochondrial DNA (1.1%) analyses indicate a male-biased influx, pointing to a predominantly paternal contribution. This observation calls for a reevaluation of the commonly held assumption that Sarracinic rule alone accounts for North African ancestry in Iberia. Analysis of polygenic risk scores for common diseases (including breast and ovarian cancer, Alzheimer’s, schizophrenia, and type 2 diabetes) reveals distinct micro-geographical patterns of disease risk in the region, stratified by genetic clusters. These insights highlight the importance of further research into implications for public health policy.

Introduction​

There is a growing interest in characterizing genome-wide variability and genetic structure at small or regional geographic scales in different human populations. Several national European examples include extensive GWAS studies performed in Britain (Leslie et al., 2015), France (Karakachoff et al., 2015), Finland (Kerminen et al., 2017), Italy (Fiorito et al., 2016), Ireland (Byrne et al., 2018), Scotland (Gilbert et al., 2019), and Spain (Bycroft et al., 2019). This renewed interest in fine-scale genetic structure is essential for controlling genetic ancestry in common variant association studies (Mathieson & McVean, 2012), as well as understanding rare variant association studies, since rare variants are often specific to certain geographically clustered populations (Genomes Project et al., 2015). In addition, fine-scale genetic structure can clarify the relationships between closely related populations and uncover recent historical events. However, there is a scarcity of studies performed with sequencing data, since most of these population studies have been carried out using genome-wide genotypic data; an early notable exception being the Genome of the Netherlands Project (Genome of the Netherlands, 2014). Large-scale sequencing is far more precise and informative in revealing genetic structure, and it also allows for novel variants to be discovered. In the Iberian Peninsula, the shortage of large-scale genomic studies is even more noticeable, with, to our knowledge, only three studies published, one using Whole Exome Sequencing (WES) data (Dopazo et al., 2016), and the other two using Whole-Genome Sequencing (WGS), one of them focusing in the northeastern Iberian region of Catalonia (Valls-Margarit et al., 2022), and the other primarily addressing signals of natural selection (Garcia-Calleja et al., 2025).
Galicia is a region occupying the northwestern corner of the Iberian Peninsula that possesses several unique characteristics. Its geographical isolation from the rest of the peninsula has, to a certain degree, maintained its identity and distinctive culture, including its own language. Besides, at the genomic level, Galicia seems to show a special differentiation, with examples of possible founder effects related to several recessive Mendelian diseases such as ichthyosis (Esperón-Moldes et al., 2019) or Wilson disease (Brage et al., 2007), as well as exhibiting signs of local genetic inbreeding (Esperón-Moldes et al., 2019) or isolation and low genetic diversity (Salas et al., 1998). Its peripheral location in Europe and its relative isolation make Galicia particularly intriguing for studying population genetic structure. Recent data on the Spanish population by Bycroft et al. (2019) has reported extraordinary fine-scale population substructure within Galicia, albeit using genotypic data (∼693K SNPs) and exhibiting a geographically biased sampling towards the southwestern region of the Pontevedra province, where the great majority of the Galician samples were recruited.
Previous studies have reported a significant North African genetic contribution to Iberian populations, particularly in Galicia, believed to stem primarily from Arab-Berber migrations during the Muslim conquest of the Iberian Peninsula in 711 CE. Sarracinic rule lasted in various forms until the fall of Granada (in Andalusia, in the southern part of the peninsula) in 1492. The Kingdom of Asturias, which encompassed present-day Galicia, served as a key Christian stronghold against Muslim expansion. Over time, this resistance evolved into the Reconquista, a prolonged series of military campaigns aimed at reclaiming territories under Sarracinic rule. By the mid-8th century, Galicia had largely returned to Christian control. Therefore, Galicia played a pivotal role in the broader Reconquista efforts as this region served as a refuge for Christian elites and populations, and from there, the first resistance movements began. While the direct impact of Muslim rule in northern Iberia, including Galicia, was relatively limited compared to other regions, genetic studies suggest a substantial proportion of Galician ancestry originated from North Africa. The availability of WGS data now provides an unprecedented opportunity to investigate this ancestry with higher genetic resolution than previous studies.
In this study, we generated WGS data from 94 individuals of Galician origin, a sample size comparable to that of other WGS studies conducted in Iberia and Europe. The availability of high-resolution genomic data enabled us to investigate inbreeding patterns and explore the extent of African and Middle Eastern admixture in Galicia. We analyzed genetic variation in the region at the highest level of resolution to build upon previous remarkable findings suggesting the presence of ultrafine population structure in Galicia (Bycroft et al., 2019). To further evaluate these findings, we additionally genotyped samples from across Iberia (n = 453), closely replicating the sampling strategy of Bycroft et al. (2019) to critically assess their reported results in the Galician population and, more broadly, in Spain. Finally, we conducted the first investigation into the biomedical implications of fine-scale population structure within Galicia, analyzing micro-spatial patterns of polygenic risk for various common diseases.

Material and Methods​

Sampling​

We collected 94 DNA samples (hereafter referred to as the GALOMICS [GAL] dataset) from donors representing the entire territory of Galicia (northwest Spain), ensuring maximum geographic homogeneity in the sampling process. Of these, 91 donors passed all the filters for subsequent analysis (see below). The average distance between the donors’ birthplaces and those of their four grandparents was 6.6 km (SD: 15.8), with a median of 0 km (IQR: 0–5.5 km). The average distance among the birthplaces of the four grandparents was 5.4 km (SD: 10.7), also with a median of 0 km (IQR: 0–5.2 km). No missing data were recorded regarding the biogeographical origins of the donors or their grandparents. The sampling primarily represents a rural population, defined as areas with fewer than 10,000 inhabitants (according to official statistics: IGE (Instituto Galego de Estatística)). Specifically, the median population size of the donors’ birthplaces was 9,080 inhabitants, with 59.3% of donors born in rural areas as defined above. Additionally, the median population size of the birthplaces of the donors’ grandparents was 5,168 inhabitants. In total, 63.7% of donors have all four grandparents born in rural areas, 75.8% have at least two grandparents from rural areas, and only 20.8% have all four grandparents born in urban areas. Figure S1 – Supplementary File. By province, the samples were distributed as trinc: 39.6% from A Coruña, 20.9% from Ourense, 23.1% from Lugo, and 16.5% from Pontevedra.
Saliva samples were collected using Oragene OG-500 devices (DNA Genotek), and DNA was then isolated using the prepIT L2P kit from DNA Genotek, trinc the standard protocol. Written informed consent was obtained from all Galician donors before the research. The rights of participants were safeguarded during the research, and their identities were protected. The study complies with all applicable Spanish regulations, including the Biomedical Research Act (14/2007-3 of July), the Autonomy of the Patient Act (41/2002), Decree SAS/3470/2009 for observational studies, and the Data Protection Act (15/1999). The experimental protocol was approved by the Galician Ethics Committee (registration code: 2017/399).
In addition to performing WGS on the GAL dataset, we genotyped DNA samples from the Spanish National DNA Bank (Presentación | DNA National Bank; hereafter referred to as the NDNAB dataset). These samples were collected from 453 donors representing various Spanish provinces across different autochthonous regions, including all four provinces of Galicia. Donors in this dataset were selected to ensure broad regional representation in the Spanish territory, somewhat comparable to the sampling strategy used by Bycroft et al. (2019); samples from the Spanish National DNA Bank were, in fact, also used by these authors.

Reference Populations​

We used three different repositories for WGS data of reference populations. From the 1000 Genomes Project (1000G; ), we accessed data representing major European, Asian, and African populations. Variant files were initially downloaded from the 1000G public FTP site (). For the bioinformatic processing of the 1000 Genomes data (Index of /vol1/ftp/phase1/data), we leveraged previous bioinformatic developments (Amigo et al., 2011; Amigo et al., 2008). From the Human Genome Diversity Project (HGDP), we used data from seven datasets representing additional whole genomes from Europe, North Africa, and the Middle East (Cann et al., 2002; Cavalli-Sforza, 2005). The downloaded data from the 1000G contained 78,090,318 SNPs, while the HGDP dataset included 64,214,739 SNPs. The combined overlapping data from all these population datasets (GAL + 1000G + HGDP) consisted of 11,085,935 SNPs. To better understand the African contribution to the Galician genomes, we enriched the admixture analyses by incorporating North African and Middle East WGS data from Serra-Vidal et al. (2019). A summary of the datasets used in this study is provided in Table 1, and Figure 1A shows a map of the datasets included.

(imagen visible para usuarios registrados)

Whole-Genome Sequencing​

Sequencing was conducted by BGI using the trinc protocol. Genomic DNA samples were randomly fragmented using Covaris technology, producing fragments of approximately 350bp. After fragment selection, DNA fragments underwent end repair, trinc by the addition of an “A” base to the 3’ end of each strand. Adapters were ligated to both ends of the end-repaired/dA-tailed DNA fragments. This was trinc by ligation-mediated PCR (LMPCR) amplification, single-strand separation, and cyclization. Rolling circle amplification (RCA) was then performed to create DNA Nanoballs (DNBs). The qualified DNBs were loaded onto patterned nanoarrays, and paired-end reads were generated using the BGISEQ-500 platform. High-throughput sequencing was performed for each library to ensure that the average sequencing coverage requirement for each sample was met. The sequencing-derived raw image files were processed using the BGISEQ-500 base-calling software with default parameters to generate the sequence data as paired-end reads. This sequence data, referred to as “raw data,” was stored in FASTQ format.

Genotyping of the Spanish DNA National Bank​

Samples of the NDNAB dataset were genotyped using the Axiom® Genome-Wide Human Origins 1 Array at the Centro Nacional de Genotipado (CEGEN) in Santiago de Compostela, Spain. Trinc sample processing, genotype calls and quality control metrics were generated using the Affymetrix Genotyping Console™ (GTC) 4.1.2 software, adhering to standard commercial guidelines. As part of this process, sample dish QC (DQC) values were calculated, and any sample with a DQC value below the default threshold of 0.82 was excluded from the study. Genotyping was performed on the remaining high-quality samples using the “AxiomGT1_all” algorithm, incorporating all SNPs available on the array. In total, 599,424 SNPs were successfully retrieved from the NDNAB dataset.

Bioinformatics Analysis​

Quality control (QC) of DNA FastQ was performed using FastQC v0.11.9 [1] and MultiQC v1.10 (Ewels et al., 2016). They were mapped to Human Genome v37 obtained from (Google Cloud Platform) using BWA 0.7.17-r1188 (Li, 2013; Li & Durbin, 2010). PCR duplicates were marked using Picard 2.27.0 (Picard Tools - By Broad Institute). Variants were discovered using Google Deepvariant 1.5.0 (Poplin et al., 2018) and GLnexus (Yun et al., 2021). The Combined Annotation Dependent Depletion (CADD) GRCh38-v1.6 dataset (Rentzsch et al., 2021) was used to obtain predictive information quantifying the deleteriousness of SNPs and small insertion-deletion variants (indels) across the human genome. This database integrates a wide range of annotations to assess the potential impact of variants, encompassing both coding and non-coding regions. CADD provides two main types of scores: raw scores, which reflect the unscaled deleteriousness of a variant based on the CADD model, and PHRED-scaled scores (used in the present study), which are log-transformed and adjusted for easier interpretation. Additionally, the dataset includes information on the functional classification of variants, such as whether they are non-synonymous, synonymous, stop-gained/lost, intronic, or fall into other categories such as regulatory regions or splice sites.

Analysis of Uniparental Markers​

Information on mitochondrial DNA (mtDNA) and Y-chromosome variation in GAL was extracted from the WGS data. MtDNA variation was contrasted against the rCRS, and haplotypes were classified into haplogroups using HaploGrep software (Weissensteiner et al., 2016). To infer Y-chromosome haplogroups from WGS data, we used the Python-based tool Yleaf (Ralf et al., 2018).

Biogeographical analysis​

To investigate clusters of genome variation, we first carried out Multidimensional Scaling (MDS) analysis using pairwise Identity-By-State values of the GALOMICA sample and the reference populations. MDS was performed using the function cmdscale (library stats) from R (). Average Identity-By-State distances were computed as the average of the matrix of identity-by-state distances computed between individual genomes against all the individuals from each reference population.

Admixture Analysis​

Maximum likelihood estimation of individual ancestries from multi-locus SNP data was undertaken using ADMIXTURE (Alexander et al., 2009); this software uses a maximum likelihood estimation of individual ancestries from multi-locus SNP data. For the analysis, we used the trinc filtering criteria: MAF > 0.05 and LD pruning with a window size of 100bp, a step size of 10, and r2 > 0.5, trinc the same approach used to compute inbreeding coefficients (Chang et al., 2015), as detailed below. We incorporated various sources of worldwide populations from the 1000G and the HGDP. First, we run ADMIXTURE in an unsupervised mode to gain a general view of the potential demographic influence of Africa on the GAL dataset. Shared ancestry between Mediterranean populations from Europe and Africa could lead to an overestimation of North African and Middle Eastern ancestry in Galicians. To address this, we also used supervised ADMIXTURE analysis using a minimal set of reference samples representing the most likely African-origin surrogates for Galician ancestry with the datasets available. In this supervised analysis, we included only genomes in which >80% or 90% of their genetic makeup was assigned to the respective North African or Middle Eastern components, thereby minimizing the impact of shared genetic influence with Europeans.
We computed f3–statistics to assess whether the genetic relationships among the three populations could be explained with or without admixture. Specifically, we used the three-population test in the form f3(XEUR,XNAF/SSH;GAL), where (XEUR) represents a particular European population subset, XNAF/SSH denotes a specific North African or Sub-Saharan African population sample, and GAL corresponds to the Galician dataset. This test evaluates whether the genetic makeup of Galician genomes can be attributed to admixture between European and North African populations. A significantly negative f₃-value (typically with Z-score ≤ –3) is strong evidence that GAL has experienced admixture between populations related to XEUR and XNAF/SSH. A positive f₃-value or non-significant Z-score suggests no detectable admixture, or that the populations chosen (XEUR, XNAF/SSH) do not adequately model the admixture sources. To ensure the robustness of our conclusions, we tested multiple combinations of European and African datasets, capturing the genetic diversity within these regions.

Dating Admixture​

We used ALDER (Loh et al., 2013) to estimate the admixture timing of a single source population from North Africa (represented by Mozabites) and Middle East (represented by Bedouins) into our recipient population, Galicians (GAL), by analyzing the LD decay within the recipient group: When a population receives genetic input from an external group, the introduced alleles initially exhibit long-range LD, which gradually decays over generations due to recombination. Input files for ALDER were prepared using EIGENSOFT (Patterson et al., 2006). The ALDER framework assumes random mating in the target population trinc the admixture event and is optimized for detecting relatively recent admixture, whether arising from a single pulse or multiple waves of gene flow (Moorjani et al., 2011).
We used fastGLOBETROTTER trinc the author’s recommendations (Wangkumhang et al., 2022) to independently infer and date admixture events. As a first step, we used CHROMOPAINTER (Lawson et al., 2012) to generate haplotype-sharing profiles by modeling each individual’s genome as a mosaic of segments copied from a set of reference populations. These haplotype profiles were then analyzed by fastGLOBETROTTER, which examines the distribution and decay of shared haplotype segments across the genome, a signal shaped by historical recombination. By assessing how ancestry contributions from different source populations vary with genetic distance, fastGLOBETROTTER infers the number, timing, and ancestral composition of admixture events. Unlike methods that rely solely on allele frequency differences, fastGLOBETROTTER derives its high resolution by using haplotype structure, which enables it to detect even subtle or ancient admixture signals with greater sensitivity. In this study, we applied fastGLOBETROTTER specifically to detect and characterize admixture signals involving North African and/or Middle Eastern ancestral sources, similar to our use of ALDER. To simplify model complexity, and given the limited availability of WGS reference datasets, our model included representatives for four broad ancestry components: sub-Saharan variation (Yoruba from the 1000 Genome Project), European ancestry (CEU from the 1000G), North African (Mozabites from the HGDP), and Middle Eastern (Bedouin from the HGDP).
To reduce the confounding effects of European-like variation in African populations, we excluded Mozabites and Bedouin individuals with <80% African-like ancestry. Model comparison indicated that a two-date admixture model did not improve the fit relative to a single-date model. Therefore, we report results from the simpler, single-date admixture scenario, which we consider the most parsimonious representation of the underlying demographic history. We also report 95% confidence intervals for admixture dates based on 200 bootstrap resamples, providing robust statistical uncertainty.
Despite the low-resolution nature of the NDNAD dataset and its limited sample size, we also aimed to explore temporal signals across different geographic regions, to evaluate whether there is evidence supporting differential entry dates of North African and Middle Eastern ancestry into the Iberian Peninsula. In particular, we focused on possible differences between the northern regions and the central/southern parts of the Peninsula. This question arises from the widespread assumption that the Arab rule, beginning in 711 CE and trinc by the gradual Christian Reconquista, was the primary source of North African ancestry in Iberia. However, there is limited historical support for a strong Arab-Berber presence in the far North of the Peninsula, where the Muslim political and military influence was more restricted and often transient. Consequently, while the Sarracinic conquest could plausibly account for much of the North African genetic signal in central and southern Iberia, the presence of this component in the North may reflect alternative demographic processes and earlier or later migratory events, potentially including Roman-era, pre-Sarracinic, or post-Reconquista population movements.
fastGLOBETROTTER also provides estimates of ancestry proportions from the reference populations considered (e.g. Mozabites and Bedouins), which are expected to be consistent with the proportions inferred using ADMIXTURE.
We used the latest estimate of generation time by Wang et al. (2023), which provides one of the most comprehensive and robust empirical estimates of human generation intervals, based on a genomic approach using de novo mutations (DNMs). According to these authors, for a typical European population over the past 40,000 years, the sens-averaged generation time is estimated at 26.1 years.

Analysis of FineSTRUCTURE​

Clusters of individuals in the Galician dataset were inferred using fineSTRUCTURE (Lawson et al., 2012) trinc the author’s recommendations. The number of clusters is not determined a priori; it is just the number estimated under the underlying probability model. We used the CHROMOPAINTER package (Lawson et al., 2012) to estimate the coancestry matrix derived from the clusters estimated by the software and to build a phylogenetic tree of clusters. FineSTRUCTURE was first run on the GAL dataset to investigate population clusters within the Galician territory. To extend the analysis to a national scale and contextualize Galicia within Spain, we analyzed the clustering pattern in the NDNAB dataset.
Principal Component Analysis (PCA) analyses based on the coancestry matrix were generated using R scripts provided by the authors (available at The world's most powerful ancestry inference software).

Analysis of Runs of Homozygosity and Inbreeding​

Runs of homozygosity (ROH) and inbreeding were assessed in GAL and other reference populations using WGS data analyzed in PLINK (Chang et al., 2015). ROH segments were identified trinc previous protocols (Serra-Vidal et al., 2019), applying sliding windows of 50 SNPs across the genome, defining segments containing a minimum of 50 SNPs within 100 kbp, allowing gaps of up to 2000 kbp between consecutive SNPs, with no more than one heterozygous site and five missing sites. To evaluate differences between GAL and other populations, Wilcoxon tests were computed on multiple ROH metrics: total counts of short and long ROH segments (nROHS and nROHL; respectively), cumulative length of short and long ROH segments (sROHSand sROHL, respectively), and average length of short and long ROH segments (mROHS and mROHL, respectively); here, short segments are defined as those ≤1.5 Mb and long segments as those >1.5 Mb. We also computed FROH, which estimates the total inbreeding coefficient (estimator of FIT of Wright’s fixation index (Hartl & Clark, 2007)) and is defined as the sum of ROH segments >1.5 Mb divided by the total length of the genome in Mb (in our study, 2,795 megabases [Mb]).
We additionally computed four single-point estimators of inbreeding coefficients based on allele frequencies. The coefficient FHET measures the average SNP homozygosity within an individual relative to the expected homozygosity of alleles randomly drawn from the population (genome-wide homozygous excess due to inbreeding). Fhat1 represents the variance-standardized relationship minus 1 (equivalent to the diagonal of the covariance matrix used for PCA). Fhat2is based on the homozygous excess. The estimator FHET is analogous to the FHAT2 estimator, both measuring the excess of homozygosity, but they are not identical; while FHET is a ratio of sums, FHAT2 is a sum of ratios (Gazal et al., 2014). Fhat3 is based on the correlation between uniting gametes; it uses the initial definition of the inbreeding coefficient proposed originally by Sewall Wright in 1922. PLINK v1.9 (Chang et al., 2015) software was used to curate the data and perform most of the analyses. R statistical software (The R Core Team, 2019) was used for the fruta of identity-of-state values used in the MDS analysis, ROH, and F inbreeding metrics, and graphic representations.

Exploring Variants of Biomedical Interest​

We are interested in exploring genomic variation in Galicia on a WGS scale, particularly its biomedical relevance compared to the broader Iberian population. This could help anticipate the potential benefits of conducting large-scale whole-genome projects at a regional level, rather than focusing solely on, e.g. exome variation, which, while more cost-effective, may not capture the full spectrum of relevant genetic diversity.
First, we conducted a single-point allele association test, comparing single-nucleotide polymorphism (SNP) variants detected in GAL with those in IBS (n = 107) and a merged European dataset (CEU + GBR + IBS + TSI; n = 404). This approach mimics a GWA’s study, where typically a cohort of cases (such as GAL) is compared to a cohort of controls (such as IBS) (Salas et al., 2017). For this analysis, we included only variants with a minor allele frequency (MAF) above 0.05 in cases and controls and a Hardy-Weinberg equilibrium P-value > 0.001 for GAL, IBS, and the European merged dataset.
Second, to investigate the impact of accumulated pathogenic variants in genes that could increase susceptibility to common diseases in Galicia compared to other Iberian/European populations, we performed a statistical association test using the collapsing method known as the Optimized Sequence Kernel Association Test (SKAT-O) (Lee et al., 2012). This method combines a set of generalized SKAT tests with varying proportions of SKAT and Burden. We incorporated annotated information for each variant obtained from the Combined Annotation Dependent Depletion (CADD) v1.6. database (Rentzsch et al., 2021). We undertook this gene-based statistical analysis using the SKAT R library (Lee & Miropolsky, 2023) and considering all variants within genes, aiming to capture rare variations that could be of biomedical interest and might go unnoticed in single-point association tests. Genes with at least two variants were considered for these analyses. A nominal significance level was set to 0.05. To monitor multiple tests, we used the Bonferroni correction.
The top SNP and gene candidates from these analyses were further investigated for their association with medical conditions. For this, we gathered information from SNPedia, (downloaded from Category:Is a snp - SNPedia [for SNP variation] and Category:Is a gene - SNPedia [for genes], updated in October 2023) and classified it into ten macro-areas (October 2024). The SNPedia contains information on 111,702 genetic variants and 2,164 genes (SNPedia:FAQ - SNPedia Pedia.3F).
Third, given the growing interest in predicting genetic predisposition to diseases and traits within the biomedical community, we analyzed potential regional differences in Polygenic Risk Score (PRS) predictions, comparing PRS values between the GAL and IBS populations. PRSs estimate disease risk based on the cumulative effect of multiple genetic variants, typically inferred from large-scale GWAs. To streamline our analysis, we selected PRSs for common diseases with significant public health relevance, including five genome-wide PRSs from Khera et al. (2018): i) Coronary Artery Disease (PRSCAD; PGS000013): 6,630,150 polymorphisms; 6,476,712 retained after merging GAL and IBS; ii) Atrial Fibrillation (PRSAF; PGS000016): 6,730,541 polymorphisms; 6,562,700 retained; iii) Type 2 Diabetes (PRST2D; PGS000014): 6,917,436 polymorphisms; 6,715,496 retained; iv) Inflammatory Bowel Disease (PRSIBD; PGS000017): 6,907,112 polymorphisms; 6,712,364 retained; and v) Breast Cancer (PRSBC1; PGS000015): 5,218 polymorphisms; 4,532 retained. Additionally, we included: i) a second Breast Cancer PRS metric (PRSBC2; PGS000004), implemented in the CanRisk calculator 🙂 313 polymorphisms, 286 retained; ii) Ovarian Cancer (PRSOCAC; PGS003394): 36 polymorphisms; 36 retained (Lee et al., 2019), also implemented in CanRisk, iii) Acute Lower Respiratory Infection (PRSALRI; PGS000925): 402 polymorphisms, 346 retained; the only infectious disease PRS indexed in the PGS Catalog to date, included due to various EWAS carried out on infectious diseases in the Galician population to date (Camino-Mera et al., 2024; Pardo-Seco et al., 2024; Salas et al., 2018), iv) Schizophrenia (PRSSCH; PGS000133): 604,645; 602,522 retained (Zheutlin et al., 2019), v) Autism Spectrum Disorder (PRSASD; PGS000327): 35,087 polymorphisms; 34,362 retained (Grove et al., 2019), and Alzheimer’s Disease (PRSALZ; PGS004146): 915,771 polymorphisms; 914,409 retained (Monti et al., 2024). This selection allows for the exploration of PRSs composed of a wide range of polymorphisms, from a few dozen to several million, capturing different levels of genome-wide architectures. Notably, ∼98% of the variants included in these PRSs are in non-coding regions of the genome, which would not be detected in EWS studies. While these non-coding variants do not alter protein sequences, they can influence gene expression and contribute to disease susceptibility. PRS calculations were performed using PLINK (Chang et al., 2015), and values were standardized to a unit standard deviation (SD) within the GAL + IBS pooled dataset.
In addition, we have developed a web tool () that enables users to search for specific variants in the GALOMICS whole genome database, which contains information on over 18 million biallelic variants. In the “Search” tab, users can query genetic variants by two main criteria: by rsID (e.g., rs1805005) or by a combination of chromosome and position (e.g., 16: 89919436). The trinc visualizations are provided: i) A geometric map displaying the coordinates of the samples, including layers for individuals as well as their parents and grandparents., ii) Various charts indicating the samples in GAL that carry the reference allele vs. those with the alternative allele, showing the genotypic composition of individuals, iii) A representation of allele variant frequencies across different populations worldwide, and iv) Biomedical data of interest. For graphical representations of allele frequencies, data is sourced from external databases, including: i) Ensembl (); Entrez: ClinGen: Allele Registry? caid=; UniProtKB: UniProt; OMIM: Home - OMIM - NCBI; dbSNP: Home - SNP - NCBI. The data obtained from these sources is structured in JSON format. Additionally, a ClinVar Data Table displays the identifiers assigned to the queried variant by ClinGen, UniProtKB, OMIM, and dbSNP. The table includes details such as the cDNA Change (alteration at the complementary DNA level), the variant’s position in the hg38 version of the human genome, the associated Trait Name, the Trait Reference, and the Protein Change (protein alterations caused by the variant).

Visualization of Genetic Data in Geographic Maps​

To visualize the biogeographical distribution of genetic cluster frequencies, we first calculated individual cluster frequencies at each sampling location. These frequencies were then spatially represented on geographic maps using SAGA v. 9.6.1 (SAGA - System for Automated Geoscientific Analyses) (Conrad et al., 2015) and the ordinary Kriging interpolation method. For the GAL samples, we had detailed information on the specific birthplaces of donors and their grandparents. In contrast, for the NDNAB samples, data resolution was limited to the provincial level. Therefore, for practical purposes, each data point was assigned to the capital city of the corresponding province. Consequently, each point in this case represents all individuals from that province, sharing the same geographic coordinates.

Results​

Overview of Galician WGS Variation​

We first removed non-autosomal variants, indels, and variants with a genotyping rate below 99%. After applying additional filters (e.g., removal of triallelic SNPs and indels), the dataset was reduced to over 17.2 million variants, which were used for downstream analyses (unless additional filtering was required, e.g., for Hardy-Weinberg disequilibrium, linkage disequilibrium [LD]). Approximately 36.4% of the variants were located within intronic regions, while 32.3% were in intergenic regions. Additionally, 15.2% of the variants were identified in regulatory or splice site regions, and 0.8% were in exons. Among the variants located in exons, >35K were non-synonymous, >29K were synonymous, and >79K were found in 3’/5’-untranslated regions (UTR). The remaining 500 variants were classified as either stop-gained or stop-lost.
We first investigated the familial relationship between donors in GALOMICS as done previously (Camino-Mera et al., 2024; Pardo-Seco et al., 2016). Of the initial 94 WGS, three were excluded due to close relationships (Figure 1B) or their behavior as outliers in the MDS analysis (see below).
Finally, we generated a web tool, GALOMICS (), to facilitate the search for allele variants in the GAL database. The tool has been designed to be scalable, anticipating the integration of new variants from other datasets. Many of these datasets represent disease conditions in Galician donors, some of which have already been generated, e.g. (Butler-Laporte et al., 2022; Camino-Mera et al., 2024; Pardo-Seco et al., 2024; Salas et al., 2018; Salas et al., 2017).

Biogeography and Admixture Analysis of Galician Genomes​

As expected, the MDS plot (Figure 1C) of worldwide population genomic datasets shows most samples clustering around the three main continental groups: sub-Saharan Africa, Asia, and Europe. The first dimension (accounting for 12.35% of the variation) primarily separates the sub-Saharan group (LWK and YOR) from the rest, while Dimension 2 (7.24% of the variation) distinguishes Central Asians (CHB and JPT) from Europeans (represented by BAS, CEU, FRA, GAL, GBR, IBS, SAR, and TSI) and sub-Saharan Africa. The North African datasets (BED, DRU, MOZ, and PAL) are more closely related to the European cluster, slightly shifted towards the sub-Saharan vertex, with a few exceptional profiles very closely related to sub-Saharans. In a close-up view of the European/North African cluster (Figure 1C; left), Galician genomic profiles are embedded within the European cluster but are not positioned at its core. Instead, they show a discrete shift towards the North African pole. Dimension 3 (0.64%; Figure 1C; right) separates North Africans from Europeans, with Bedouins (BED) at the opposite pole from Europeans, while Dimension 4 (0.29%) separates North Africans and Europeans from Mozabites (MOZ; Figure 1C; right). These two dimensions highlight Bedouins and Mozabites at the extremes of North African variation, with the rest of the North African samples showing more genetic affinity to Europeans.
We used the unsupervised clustering algorithm of ADMIXTURE to estimate ancestry proportions in Galician genomes, which are influenced by gene flow between African and European populations (Figure 2A). Our analysis utilized population datasets representing major continental and potentially ancestral groups, including sub-Saharan Africa, North Africa, Asian, and Europe. We enriched this analysis with other North African samples analyzed in Serra-Vidal et al. (2019). We explored values of K ranging from 2 to 10 ancestral contributors, but we only discuss results for K = 3 and 5, because these are relevant in the present demographic context. At K = 3, ADMIXTURE identifies ancestry components originating from three main geographic locations, sub-Saharan Africa, Central Asia, and North Africa/Europe. When K = 4, the model consistently distinguishes the North African component from the primary European one, suggesting a shared ancestry between Europe and North Africa (Figure 2A; left). At K = 5, which corresponds to the model with the lowest cross-validation value in the admixture analysis, the algorithm further divides the North African/Middle East component into two subcomponents. One of these subcomponents mainly characterizes North African Mozabites (hereafter referred to as ‘Mozabites North African’-like ancestry), while the other reflects the variation mainly observed in the Middle East (represented by Palestinian, Bedouin and Druze) and the other North African datasets (hereafter referred as to ‘non-Mozabites North African’-like ancestry). All the datasets in North Africa and the Middle East have, however, variable proportions of these two ancestries. Moreover, these two North African ancestries are also present in other European datasets FRA, SAR, TSI, IBS, and GAL, while the ‘non-Mozabite North African’-like ancestry is absent in the BAS, and both North African ancestries are absent in CEU and GBR.

(imagen visible para usuarios registrados)

To better estimate the North African/Middle East component in GAL, we performed a supervised ADMIXTURE analysis using the trinc reference populations (Figure 2B, right): YRI and LWK to represent the sub-Saharan African component, CHB and JPT for Central Asia, and GBR and CEU for Europe. For the North African and Middle East dataset, we selected only those individuals having >90% (in an initial run) or >80% (in a second run) of their primary ancestry component, representing their most genuine North African and Middle East ancestry. This filtering aimed to minimize the impact of their European ancestry on the estimates. Under the 90% threshold, only a few Bedouin genomes (n = 18) met the criteria, while a higher number of Bedouin (n = 21) and several Mozabite genomes (n = 18) passed the 80% filter. In the initial run, GAL shared ancestry proportions of 84.5%, 13.5%, and 2.0% with Europe, Middle East, and sub-Saharan Africa, respectively, compared to 87.0%, 11.6%, and 1.2% in IBS, and 92.2%, 7.8%, and 0.0% in BAS. In the second run, GAL shared ancestry of 83.2%, 7.4%, 8.7% and 0.6% with Europe, Middle East, North Africa and sub-Saharan Africa, respectively, compared to 86.9%, 6.7%, 5.9% and 0.4% in IBS, and 94.9%, 4.0% and 0.0% in BAS. Therefore, the North African/Middle East component in GAL ranged from 13.5% to 16.1% (miccionan: 14.8%) compared to 7.8% to 12.6% in IBS (miccionan: 8.25%), and 5.1% and 7.8% in BAS (miccionan: 6.45%). The differences between ancestry components between these datasets were statistically significant for all comparisons (Figure S2 – Supplementary File). In addition, we also detected a minor yet statistically significant Sub-Saharan ancestry to GAL, ranging from 1.0%–2.8% (miccionan: 1.9%), which is higher than values observed in IBS (0.3%–1.2%; miccionan: 0.75), and BAS (0%). Interestingly, although the North African and Middle East component is very homogeneous across Galicia, we observed a subtle but clear clinal geographic pattern, with a slightly higher proportion in the southern regions, gradually decreasing toward the North (Figure 2B).
According to fastGLOBETROTTER, the estimated admixture contribution from Mozabite (North African-like) ancestry to the GAL dataset is 6.9%, while Middle Eastern-like ancestry accounts for 4.6%, yielding a combined total of 11.5%. This value is slightly lower than the estimates obtained from ADMIXTURE.
To further test the robustness of these admixture inferences, we used the f3-statistics. As shown in Figure 2C, the results of the f₃-statistics indicate that the admixture patterns observed in the GAL dataset are statistically compatible with historical gene flow involving various combinations of European and African source populations. Specifically, several f3(XEUR, XNAF/SSH; GAL) tests yielded significantly negative values, suggesting that GAL may represent a population formed through admixture between groups genetically related to both European and African ancestries.
Finally, the analysis of mtDNA haplogroups in the GAL dataset reveals that only 1.1% of the samples belong to North African lineages (U6 haplogroups), with no Sub-Saharan lineages (L haplogroups) detected. In sharp contrast, an analysis of the Y-chromosomes of GAL males (n = 33) indicates that 21.2% (E1b1b1 haplogroups) show a contribution from North African/Middle Eastern lineages.

Time Estimates of African Admixture​

Time estimates from fastGLOBETROTTER indicate the occurrence of an admixture event in the Galician population that is best explained as gene flow between a predominantly European ancestral group and a more admixed African/Middle Eastern group. The inferred minor source shows a particularly strong North African signature, represented by Mozabites (MOZ), contributing approximately 42%, alongside a smaller Middle Eastern component, represented by Bedouins (BED), accounting for about 19% of the same source. The overall admixture event is dated to approximately 54 generations ago, which translates to a calendar range of around 510–690 CE, assuming a generation time of 26.1 years.
A two-date admixture model was also tested, but it did not provide a better fit than the one-date model, supporting the hypothesis of a single, temporally localized admixture event. The admixture signals were further characterized by the presence of a sub-Saharan African component, with YRI acting as a proxy. While this component was less prominent, it may reflect either a genuine signal of sub-Saharan input or result from model limitations in distinguishing between shared North African and sub-Saharan ancestry.
To validate and complement the findings from fastGLOBETROTTER, admixture dating analyses were also performed using ALDER. In agreement with fastGLOBETROTTER, this software estimated the admixture involving North African ancestry (using MOZ as a proxy) to have occurred approximately 50 generations ago, and admixture with Middle Eastern ancestry (using BED as a proxy) around 54 generations ago. In addition, ALDER identified admixture involving sub-Saharan African ancestry (using YRI and LWK as proxies) at approximately 57 generations ago, which corresponds to around 540 CE.
Taken together, both fastGLOBETROTTER and ALDER yield overlapping time estimates for the introduction of North African and Middle Eastern ancestry into Galicia. For interpretative clarity, we adopt the central point estimates of both methods, 52 generations from ALDER and 54 from fastGLOBETROTTER, to define a reference interval of admixture of 620–670 CE.
We also estimated the timing of the introduction of the North African/Middle Eastern genetic component into various Iberian regions using the NDNAD dataset. While these estimates must be interpreted with caution due to the dataset’s relatively low SNP density and limited sample sizes, they nonetheless reveal informative regional patterns. The inferred admixture dates (in generations, with 95% confidence intervals) were as trinc: Andalusia/Central Iberia: 33 (30–35), Catalonia: 39 (34–44), Basque Country/Navarra: 55 (30–70), Cantabria: 46 (40–61), and Asturias: 51 (45–55). Notably, Galicia showed an admixture time of 43 generations in the NDNAD dataset, compared to 52 generations when using the higher-resolution WGS data from the GAL cohort. This discrepancy suggests that the NDNAD estimate may underestimate the true admixture date due to its lower resolution. Despite the inherent limitations, the overall trend across regions supports an earlier admixture in northern Iberia compared to the South and the Center. To further validate this signal, we merged the northern regions (Basque Country/Navarra, Cantabria, Asturias, and Galicia; n = 164) to increase statistical power and compared them with the Andalusia/Central Iberia group (n = 191). This analysis yielded an admixture estimate of 48 generations (44–51) for northern Iberia, significantly older than the 33 generations estimated for the South-Central regions. These findings suggest that the introduction of North African/Middle Eastern ancestry into Iberia may have occurred in multiple waves or via regionally distinct demographic processes, with earlier influxes into the North than previously assumed.

Fine-Clustering Pattern in Galicia with Reference to National Patterns​

At the highest level of genome resolution (WGS), fineSTRUCTURE analysis of the GAL dataset identified 10 terminal branches, though only five formed clusters containing at least four individuals (Figure 3A). The Galician genetic landscape is predominantly shaped by a single major cluster, referred to here as the ‘Main’ cluster, comprising 66 genomes and uniformly distributed across most of the region (Figure 3B).

(imagen visible para usuarios registrados)

The remaining four clusters are significantly smaller and more localized. The ‘Porto do Son’ cluster (West Coast, n = 8) shows the most pronounced genetic isolation from the broader Galician population. The ‘Cuntis/A Estrada’ cluster (Inner Pontevedra, n = 6) and the ‘A Mariña/Foz’ cluster (northeast Coast, n = 4) are also geographically distinct, with the latter located along the northern coastland. Lastly, a small and scattered ‘Central/West’ cluster (n = 5) is dispersed throughout central and West Galicia (Figure 3A and 3B).

Interpolated frequency maps further emphasize the dominance of the ‘Main’ cluster, while the ‘Porto do Son’ cluster stands out distinctly (Figure 3B). The ‘Cuntis/A Estrada’ and ‘A Mariña/Foz’ clusters are less visible due to their blending with the ‘Main’ cluster in their respective areas. Meanwhile, the ‘Central/West’ cluster, being both small and dispersed, is further diluted by the influence of the dominant ‘Main’ cluster. This overall pattern highlights the remarkable homogeneity of the Galician genetic landscape, with only limited instances of localized isolation, most notably in the village of Porto do Son.

The PCA of the fineSTRUCTURE results (Figure 3C) also reveals the five clusters described above. PC1 (21.4% variance explained) places the ‘Porto do Son’ and ‘A Estrada//Cuntis’ clusters at the opposite ends of the axis, yet both align on the same pole of PC2 (15.5%), distinguishing them from the remaining clusters. PC3 (8.0%) positions the ‘Central/West Galicia’ and ‘A Mariña’ at opposite extremes, while PC4 (5.2%) further separates these two clusters from the rest. The clustering pattern shown by PC1 to PC4 does not strictly trinc the phylogenetic structure suggested by the phylogenetic dendrogram, where, for example, the deepest split occurs between the ‘Central/West’ cluster and the rest. However, it demonstrates a distinct segregation of individuals into five genetically differentiated clusters.

The low genetic structure revealed by WGS data in Galicia contrasts sharply with the highly stratified landscape reported by Bycroft et al. (2019) in their study of the Iberian Peninsula, particularly Galicia, which was based on lower-resolution genomic data. These authors reported identifying 145 genetic clusters among Spanish donors, nearly half of which (∼70) were concentrated in a small, densely populated area within the Galician province of Pontevedra. Remarkably, a closer examination of their data and maps reveals that Galicia accounts for 107 of the 145 clusters (Pardo-Seco et al., 2025). Although this specific area of Pontevedra was more densely sampled in Bycroft et al. (2019) than in the GAL study, the observed discrepancy in stratification levels does not appear to be solely attributable to differences in sampling density (Pardo-Seco et al., 2025).

In light of the sharp contrast in clustering patterns, we performed a fineSTRUCTURE analysis on the NDNAB dataset (n = 453, with 474,761 SNPs after applying the same filters as for the whole-genome data). This dataset includes independent Galician samples alongside donors from various Spanish regions, enabling a detailed assessment of Galicia’s population structure within a national context. This dataset encompasses individuals from Galicia, Asturias, Navarra, Cantabria, the Basque Country, Catalonia, Andalusia, Castilla-La Mancha, and Castilla y León, all regions also examined in Bycroft et al.’s study (2019); Figure 4A. Consistent with our findings in the GAL dataset, fineSTRUCTURE identified only three clusters among Galician donors in the NDNAB dataset. Similarly, one cluster contained most of the Galician samples (n = 46), mirroring the ‘Main’ GAL cluster, while the remaining two minor clusters comprised 7 and 3 individuals, respectively. In the broader NDNAB Iberian dataset, a total of 23 clusters were detected (including the Galician ones), each closely reflecting the geographic origin of the donors (Figure 4B; Figure S3 – Supplementary File). Unlike Bycroft et al., (2019) the first major split in the phylogenetic tree of our NDNAB Spanish dataset separates clusters exclusively found in the northern and geographically adjacent region of Cantabria (three clusters) and the Basque Country/Navarra (two clusters) from the rest of the Iberian Peninsula. The next phylogenetic subdivision separates Asturias (six clusters, with one dominant cluster dominating most individuals in the region [n = 36]), trinc by Catalonia (with a unique and well-differentiated cluster; n = 94). Galician genomes emerge at the most recent branching point of the phylogeny, clustering alongside a mixed group of individuals from Andalusia and Central Iberia. This group comprises seven clusters, with three major ones including 128, 35, and 32 individuals, respectively. This clustering pattern differs markedly from that reported by Bycroft et al. (2019), where Galicians from Pontevedra, their hyper-highly stratified Galician province, exhibited the earliest genetic differentiation and accounted for about half of the total clusters found in Iberia. In their analysis, the next major split separated the Basque Country individuals from the rest of the Iberian Peninsula, trinc by a division isolating the remaining Galician samples in their dataset.

(imagen visible para usuarios registrados)

The PCA analysis of the fineSTRUCTURE clusters found in the NDNAB database (Figure 4C) further supports the phylogenetic structure observed in the dendrogram (Figure 4B). The primary separation in PC1 (8.3%) distinguishes the Cantabria/Basque Country/Navarra clusters from the rest of the dataset. PC2 (4.1%) further differentiates the Basque Country/Navarra from the remaining regions. PC3 (2.7%) highlights the genetic distinction between Asturias and Cantabria, while PC4 (2.1%) again emphasizes the differences among these northern Iberian regions. Notably, all four principal components successfully capture and distinguish the clusters identified in the dendrogram, reinforcing the observed population structure.

Inbreeding Patterns in GALOMICS​

Analyzing inbreeding levels and runs of homozygosity (ROH) segments can reveal events of demographic isolation and offer insights into a population’s demographic history and genetic diversity. This study presents the first investigation of ROH patterns in the Galician population, enabling comparisons with other European and non-European groups at the highest level of genetic resolution possible, namely, that provided by WGS data.
The Galician population (GAL) exhibits ROH patterns comparable to those observed in other European populations (Figure 5A). However, a more detailed analysis reveals slightly higher levels of inbreeding compared to the national Spanish dataset represented by the Iberian Peninsula (IBS). Specifically, GAL has a larger count of long ROH segments, longer miccionan of these segments, and larger cumulative length compared to IBS (nROHL [GAL] = 9 vs. nROHL [IBS] = 8, P-value = 1.5×10-1; mROHL [GAL] = 2.4 vs. mROHL [IBS] = 2.1, P-value = 5.2×10-8; sROHL[GAL] = 20.6 vs. sROHL [IBS] = 15.2, P-value = 9.2×10-4); Table S2. In line with these findings, GAL also has a lower count of short ROH segments, a shorter miccionan of these segments, and a smaller cumulative length compared to IBS (nROHS [GAL] = 2504 vs. nROHS [IBS] = and 2592, P-value = 5.8×10-19; mROHS[GAL] = 0.199 vs. mROHS [IBS] = 0.203, P-value = 9.2×10-19; sROHS [GAL] = 499 vs. sROHS [IBS] = 525, P-value = 3.4×10-26); Table S2. Overall, these metrics support the observations of significantly higher inbreeding levels in GAL relative to IBS and other European datasets, except when comparing long ROH segments in historically isolated populations (Basques and Sardinians), which show higher values. Additionally, total inbreeding coefficient values indicate higher inbreeding in GAL compared to IBS (FROH [GAL] = 0.007 vs. FROH [IBS] = 0.005, P-value = 9.3×10-4) and other continental European datasets, although still lower than in Sardinians and Basques; Table S2.

(imagen visible para usuarios registrados)

n agreement with ROH values, Galicia (GAL) has higher values of FHET, FHAT1 and, FHAT3 compared to IBS (FHET[GAL] = 0.005 vs. FHET [IBS] = 0.003, P-value = 6.4×10-2; ; FHAT1 [GAL] = 0.012 vs. FHAT1 [IBS] =.0001, P-value = 2.7×10-16; and FHAT3 [GAL] = 0.008 vs. FHAT3 [IBS] =.0002, P-value = 5.2×10-9); and equal values of FHAT2 for both population (FHAT2[GAL and IBS] = 0.004); Figure 5C, Table S3. Within Europe, GAL’s metrics are intermediate, with Sardinians and Basques showing the highest homozygosity values among these datasets.
Additionally, the ‘Porto do Son’ cluster significantly contributes to the higher inbreeding values in GAL compared to IBS. For instance, metrics for long ROH segments in this cluster are substantially higher (and statistically significant) than those for the general GAL: nROHL [Porto do Son] = 11; mROHL [Porto do Son] = 3.5; and sROHL [Porto do Son] = 45). Conversely, short ROH segment metrics are generally lower: nROHS [Porto do Son] = 2442; mROHS[Porto do Son] = 0.199; sROHS [Porto do Son] = 487; Table S2. In fact, the inbreeding values for the GAL dataset excluding ‘Porto do Son’, are similar to those observed in the other clusters (GAL: nROHL [GAL without Porto do Son] = 9; mROHL[GAL without Porto do Son] = 2.6; and sROHL [GAL without Porto do Son] = 26; nROHS[GAL without Porto do Son] = 2502; mROHS [GAL without Porto do Son] = 0.199; sROHS [GAL without Porto do Son] = 497).
Overall, these results indicate that inbreeding levels in Galicia are broadly comparable to those observed elsewhere on the Iberian Peninsula, with only marginally elevated values in certain localities, such as Porto do Son.

Regional Variations of Potential Biomedical Interest​

To assess allele differences of potential biomedical relevance, we performed a single-site association test comparing the GAL cohort to the IBS dataset. This analysis was restricted to SNPs with a minor allele frequency (MAF) > 0.05 in cases and controls, as the focus was solely on common variants. A total of 85 SNP variants (out of >17M) surpassed the Bonferroni threshold (Figure S4 – Supplementary File; Table S1). Among these, only one SNP (rs9917044) was located in an exonic region (ZNF28; Zinc Finger Protein 28 gene) and was synonymous. Additionally, 25 variants were found in intronic regions, 13 were non-coding but classified as regulatory, and 46 were intergenic. When these variants were compared against a merged European cohort (CEU, GBR, IBS, and TSI) and using a MAF > 0.05 in cases and controls, 57 remained significant after applying the Bonferroni threshold. None of the 85 SNP variants appears as associated with a disease trait in SNPedia. The gene-based test between GAL and IBS identified only four genes as statistically significant after Bonferroni correction: ATP5F1AP10 (two SNPs), SDR42E1P4 (two SNPs), FRG2EP (17 SNPs), and DUX4L34 (13 SNPs); only the first two genes (supported on only two SNPs) remained significant when compared to the merged European cohort (Figure S5 – Supplementary File).

Microgeographic Stratification of Disease Risk​

By computing Polygenic Risk Scores (PRS) for eleven common diseases, we investigated whether demographic particularities in Galicia, compared to the broader Iberian (IBS) population, influence the distribution of alleles associated with disease risk in GAL genomes. For most PRS metrics, no statistically significant differences were observed between GAL and IBS. However, despite their small quantitative effect, the differences were statistically significant for PRSASD (IBS median = −0.22 [IQR: −0.68–0.53], GAL median = 0.15 [IQR: −0.41–0.71]; Wilcoxon test P-value = 0.038) and PRSIBD (IBS median = −0.25 [IQR: −0.89–0.50], GAL median = 0.25 [IQR: −0.62–0.86]; P-value = 0.007); Figure S6 – Supplementary File. The wide IQR in these metrics suggests substantial variation among individuals, indicating that some donors have markedly higher or lower polygenic risk relative to others in the group.
To further investigate the microgeographic stratification of disease risk within Galicia, we analyzed the spatial distribution patterns of PRS metrics for the eleven common diseases examined across the region. Rather than being homogeneously distributed, the maps in Figure 6A illustrate distinct geographic patterns of disease risk. Notably, ovarian cancer risk is more pronounced along the western Atlantic coastline, peaking in the southwestern coastal region. In contrast, breast cancer risk (although slightly varying depending on the PRS used) is generally elevated in both eastern and western Galicia but lower in a central North-to-East longitudinal band. Coronary artery disease and inflammatory bowel disease exhibit broadly similar geographic distributions. Atrial fibrillation risk trinc a pronounced South-to-North gradient, with lower risk in the South and higher risk in the North. Type 2 diabetes risk peaks in central Galicia and reaches its lowest values in the southeastern corner. Acute lower respiratory infection risk is lowest in the northwestern corner and highest in a small northeastern region. Autism spectrum disorder exhibits intermediate risk values from mid-central to northern Galicia, with the lowest risk in the southwest and the highest in the southeast. Schizophrenia risk is highest in the southeasternmost corner of Galicia and remains elevated across most of the region, except for the northwestern corner. In contrast, Alzheimer’s disease trinc an approximately inverse pattern.

(imagen visible para usuarios registrados)

Motivated by these findings, we further examined disease risk stratification by genetic cluster. Although the sample sizes for most clusters (except for the ‘Main’ cluster) were limited, and the different PRS scores encompass a wide range of variant counts (ranging from dozens to millions), we identified statistically significant differences in several comparisons (Figure S7 – Supplementary File). The most notable findings include: (a) Coronary artery disease risk: Significant differences were observed between ‘Porto do Son’ and ‘A Estrada/Cuntis’ (P-value = 0.016), ‘A Estrada/Cuntis’ and ‘A Mariña/Foz’ (P-value = 0.019), and ‘A Estrada/Cuntis’ and the rest of Galicia (P-value = 0.038); (b) Inflammatory bowel disease risk: Significant differences were found between ‘A Estrada/Cuntis’ and the ‘Main’ cluster (P-value = 0.026), as well as between ‘A Estrada/Cuntis’ and the rest of Galicia (P-value = 0.040). Additionally, differences were observed between the ‘Main’ cluster and ‘A Mariña/Foz’ (P-value = 0.025), and between ‘A Mariña/Foz’ and the rest of Galicia (P-value = 0.038); (c) Atrial fibrillation risk: Significant differences were detected between the ‘Main’ cluster and ‘A Mariña/Foz’ (P-value = 0.007), and between ‘A Mariña/Foz’ and the rest of Galicia (P-value = 0.008); (d) Alzheimer’s disease risk: Differences were observed between the ‘Main’ cluster and ‘A Mariña/Foz’ (P-value = 0.012), the ‘Main’ cluster and the rest of Galicia (P-value = 0.007), and between ‘A Mariña/Foz’ and the rest of Galicia (P-value = 0.027); and (e) Autism spectrum disorder risk: A significant difference was found between ‘A Estrada/Cuntis’ and ‘A Mariña/Foz’ (P-value = 0.019). Therefore, of the 13 statistically significant contrasts identified, 8 involved the ‘A Mariña/Foz’ cluster and 6 involved the ‘A Estrada/Cuntis’ cluster. Despite their relatively small sample sizes, these two clusters account for a substantial proportion of the major disease risk stratification observed in our Galician sample. In addition, PRS metrics for the different clusters can be ranked according to the diseases considered. For instance, the ‘Porto do Son’ cluster has the highest risk for ovarian cancer and the lowest for schizophrenia, the ‘A Mariña/Foz’ cluster has the highest risk for atrial fibrillation but the lowest for coronary artery disease, while the ‘A Estrada/Cuntis’ cluster has the lowest risk for atrial fibrillation and the highest risk for inflammatory bowel disease (Figure 6B).

Discussion​

Galicia’s geographic position at the far western edge of the Iberian Peninsula, combined with its historical isolation, has led some studies to hypothesize that the region may harbor unique genomic antiestéticatures (Brage et al., 2007; Bycroft et al., 2019; Esperón-Moldes et al., 2019; Reaccionariol et al., 2012; Fernández-Marmiesse et al., 2006; Salas et al., 1998). For the first time, whole-genome sequencing (WGS) data from Galicia allow an in-depth investigation at the highest possible genomic resolution, revealing patterns of genetic variation that are broadly shared with other Iberian populations, yet also display some distinct regional traits. The genetic landscape of the Galician population (GAL) reflects a demographic history consistent with early settlement, subtle differentiation from the broader Iberian Peninsula, and an early admixture with North African and Middle Eastern ancestries. Principal component analysis (PCA) and fineSTRUCTURE clustering reveal that Galicians form specific but closely related genetic clusters, showing only limited divergence from other Iberian groups.

Lack of Evidence for Long-Term Genetic Isolation​

Galicia demonstrates slightly higher values of nROHL, sROHL, and mROHL, compared to IBS and other European datasets, with statistically significant differences observed after Bonferroni correction. This pattern extends to shorter ROH segments, resulting in a higher total inbreeding coefficient (FROH) than in IBS and other Europeans. In agreement with these findings, inbreeding coefficients indicate slightly higher overall homozygosity in Galicia compared to IBS and other European datasets. While statistically significant differences with the rest of the Peninsula exist, they are quantitatively modest and largely driven by the distinct “Porto do Son” cluster (as also inferred from PCA plots and spatial genetic maps).
The slightly higher inbreeding coefficients in Porto do Son likely reflect historical geographic isolation and elevated consanguinity. Marriage patterns within small coastal or rural communities, combined with limited gene flow, can increase genetic homogeneity over generations. Nevertheless, even in Porto do Son, inbreeding levels remain within the Iberian norm, indicating that Galicia’s demographic and cultural factors produce autozygosity patterns similar to those of neighboring regions. Therefore, the slightly elevated inbreeding levels observed in Galicia challenge the hypothesis of pronounced population stratification in the region (against Bycroft et al. (2019)). If strong stratification were present, one would expect allele frequency disparities within Galicia to result in numerous genetic clusters and elevated homozygosity levels consistent with a Wahlund effect, a pattern not observed in this study. The minor divergence existing between GAL and the rest of the Iberian Peninsula most likely resulted from genetic drift and small-scale fluctuations in allele frequencies across the genome. There is no compelling evidence to suggest that long-standing geographic isolation is the primary factor shaping the genetic landscape of Galicia.
The relatively elevated African ancestry component existing in Galicia (see discussion below), greater than in other parts of Iberia, may have played a role in mitigating levels of inbreeding through increased outbreeding, thereby contributing to the preservation of genetic diversity in the Galician population.

Minimal Population Structure​

The limited number of clusters inferred by fineSTRUCTURE from Galician whole genomes (ten in total, or which only five included more than four individuals) and the moderate levels of inbreeding observed in the GAL population suggest a demographic history marked by extensive gene flow, both within Galicia and from genetically distinct, non-Galician populations, including groups from North African and Middle East. Such a pattern aligns with Galicia’s geographic and socio-historical context, characterized by: (i) the absence of major geographical barriers and (ii) a predominantly rural, dispersed demographic structure. These factors likely promoted both intra-regional mobility and external genetic exchange, helping to preserve high genetic diversity and limiting inbreeding across the relatively small geographical area of Galicia. While certain areas in Galicia may have experienced localized demographic conditions conducive to increased inbreeding or mild founder effects, as previously documented for mutations causing autosomal recessive congenital ichthyosis (Reaccionariol et al., 2012) and other inherited disorders, our data reveal a markedly lower degree of fine-scale population structure than that reported by Bycroft et al. (2019). Their study, based on low-density genome-wide data, identified ∼70 genetic clusters recognized by these authors. However, as critically reanalyzed in Pardo-Seco et al. (2025), the actual number of clusters in their data appears to exceed 107 across the Galician territory out of 145 observed in the whole Iberian Peninsula, accounting for ∼74% of the total. These clusters are heavily concentrated in a small southwestern section of Pontevedra province, near the border with northern Portugal. This strikingly high number of clusters is particularly unexpected given the province’s relatively small size, its high population density (Pontevedra is among the most densely populated provinces in Spain, with over 200 inhabitants per square kilometer; ), and its historical, cultural, and religious significance. Notably, the city of Tui in this area has long served as a key religious center in both Galicia and northern Portugal. The discrepancy is even more pronounced considering that this southwestern region is one of the most cosmopolitan areas in Galicia, home to Vigo (the province’s largest urban center and Galicia’s most populous city), and one of the world’s busiest fishing ports. Moreover, the region lacks any substantial geographical barriers that might explain deep population substructure. As critically discussed by Pardo-Seco et al. (2025), the sharp discrepancies between our results and those in Bycroft et al. (2019) could have been due to technical artifacts, amplified by the high sensitivity of the fine-STRUCTURE tool, as well as flawed misinterpretations of the data.

​

Patterns of variation in Galicia or biomedical interest​

Despite the limited statistical power, no specific common genetic markers of biomedical significance were identified that distinctly differentiate Galicia from the broader Iberian Peninsula. A gene-based analysis using SKAT, which accounts for rare variation, identified a small number of genes with statistically significant differences compared to the IBS cohort. Additionally, minor differences in disease risk between GAL and IBS were observed for various common diseases; however, the magnitude of these differences remains relatively low.
Of more interesting consideration is our observation of the disease risk micro-geographically stratified within Galicia. We demonstrate that disease risk trinc distinct spatial patterns, with certain regions exhibiting elevated risk for specific diseases while showing lower risk for others. Furthermore, for the first time, we illustrate how disease risk stratifies by genetic cluster, an approach that does not necessarily align with the patterns observed in PRS interpolated geographical maps. Notably, we show that it is possible to rank disease risk within genetic clusters, offering a novel perspective on the distribution of common disease susceptibility across the region.
While our observation of regional differences in PRS values for various conditions in Galicia is academically relevant, its implications for public health policy remain uncertain. On the one hand, PRS reflects genetic risk alone and constitutes just one of many factors contributing to overall disease susceptibility. For instance, in breast and ovarian cancer, risk stratification algorithms incorporate additional variables such as hormone replacement therapy, parity, endometriosis, contraceptive use, and age of menarche. At the regional level, where Galicia has full autonomy in managing public health policies, the primary focus would likely be on optimizing individual risk stratification through appropriate PRSs, rather than on accounting for microgeographic variation within the region. Incorporating such fine-scale differences would be particularly challenging for public health authorities due to factors such as population mobility and the complexity of tracing ancestral origins. For example, while our study includes individuals with all four grandparents born within an 80 km radius, real-world clinical cases are likely to exhibit more diverse regional ancestries. Moreover, the prevailing trend in the scientific community favors the development of PRSs with broad applicability across large populations, as seen with the widely used CanRisk model for breast and ovarian cancer. This trend is likely to hold even in populations with significantly more complex ancestry patterns than Galicia. For example, efforts are underway to develop software capable of improving PRS portability by integrating fine-scale ancestry components into PRS predictive power, as exemplified by ANCHOR, which has been applied to UK Biobank data (Hu et al., 2025).
Moreover, although subtle differences in SNP frequencies and gene markers are observed between Galicians and the broader Iberian population, the magnitude of these differences appears limited, casting doubt on the cost-effectiveness of implementing large-scale, region-specific WGS initiatives that are not explicitly focused on disease-related applications. If allele frequencies, linkage disequilibrium patterns, and genetic risk predictions remain largely consistent between Galicia and the broader Iberian region, and if existing national and international PRS models already demonstrate strong performance in predicting disease risk among Galicians, the rationale for extensive regional WGS projects becomes questionable. Instead, considering the still substantial costs associated with WGS, and even more critically, the challenges related to the storage and management of such voluminous data, a more targeted genomic approach, such as focusing sequencing efforts on coding regions through custom gene panels or WES, may offer a more efficient and cost-effective alternative. This strategy would be particularly valuable for prioritizing conditions of high public health relevance, such as hereditary breast and colorectal cancers. Supporting this view, recent research from Regeneron Pharmaceuticals (Gaynor et al., 2024) based on UK Biobank data indicates that prioritizing imputed WES over full WGS can substantially improve the discovery rate in genetic association studies, delivering critical insights while optimizing resource allocation.

Topollillo dijo:

Pero no me refiero a ponerse morenos en la playa, sino a un moreno natural, además de pelo azabache y ojos oscuros, lo que es lo mismo que el fenotipo mediterráneo por excelencia.

Observar a esta reportera gallega. Su fenotipo es una mezcla de fenotipo mediterráneo del sur entre gracile med y beride. Casos así en la tierra de los "celtas" los hay en cantidades enormes. ¿Verdad @Hermericus? No me caen bien los pueblos que se adueñan del legado de pueblos que no tienen nada que ver con ellos; genéticamente, y ya no te digo fenotípicamente, no hay ninguna diferencia entre un gallego y un murciano. Pero estos que se creen escoceses y no son más que portugueses asimilados se creen diferentes. Señores, aquí los únicos que no han cambiado nada y somos los mismos desde la edad del bronce somos los vascos.

View this content on Instagram

View this content on Instagram

Estos fenotipos no te los encuentras en Euskadi, el único refugio demográfico desde la edad del bronce. Como se puede comprobar se atisba una cantidad enorme de componente ancestral de EEF superior al 65%, posiblemente no más de un 25% de WHG, el resto claramente caucasian e iberomaurisian.

(imagen visible para usuarios registrados)

(imagen visible para usuarios registrados)

(imagen visible para usuarios registrados)

(imagen visible para usuarios registrados)

(imagen visible para usuarios registrados)

Creo que ya va siendo hora de que les quitemos a estos el mito de celtismo que nunca han tenido; yo por mi parte seguiré bromeando sobre ello porque solo me provoca risa, no hay ni una sola prueba genética que demuestre tal cosa. No son más que otros mediterráneos más.

Hacer clic para expandir...

No es por nada pero algunas de las que has puesto tienen con altísima probabilidad ascendencia amerindia. Ni siquiera parecen españolas morenas típicas.

Conclusions​

This study represents the first comprehensive analysis of WGS variation in the Galician population. Despite Galicia’s historical fruta as a genetically isolated region within the Iberian Peninsula, our results reveal no strong signatures of long-term isolation. On the contrary, we observed low levels of fine-scale population structure, challenging previous claims of extreme substructure in the area.
Importantly, we identified a significantly higher North African and Middle Eastern genetic component in Galician genomes than previously reported. This ancestry is distributed across the entire Galician territory, suggesting widespread historical admixture and sustained gene flow within the region. Such connectivity may have reduced levels of inbreeding that might otherwise be expected in a small, demographically scattered population, particularly when contrasted with other Iberian groups with larger effective population sizes. We detected a male-biased contribution to this African component, along with a subtle South-to-North frequency gradient. This geographic pattern may reflect a southern entry point into Galicia, possibly via maritime or overland routes, though further investigation is needed to confirm this hypothesis. Most notably, the timing of admixture predates Sarracinic rule in the Iberian Peninsula, challenging the traditional assumption that the observed African component in Galicia is primarily the result of Arab-Berber expansion and the subsequent Reconquista.
In addition, our analyses show that PRS vary modestly across Galicia, with evidence of microgeographic stratification both regionally and among fine-scale genetic clusters. However, we also found that WGS provides limited additional resolution over exome sequencing for the characterization of common, clinically relevant variants in this population. This highlights the importance of matching sequencing strategies to specific research goals. In many cases, targeted approaches such as exome or gene panel sequencing may offer a more cost-effective means of maximizing clinically actionable insights.

Lo que siempre dije: llegaron en masa norteafricanos en época romana tardía y en especial refugiados cristianos de la Bética más mezclados, súmale moriscos en el XVII. El noroeste fue refugio poblacional desde tiempos inmemoriales, no es de extrañar que el culto a Santiago tuviera éxito en esas tierras, eso jamás pasó en el nordeste. Si hasta de joven tenéis que oscila entre el 1-2.8%.

-Alexia- dijo:

No es por nada pero algunas de las que has puesto tienen con altísima probabilidad ascendencia amerindia. Ni siquiera parecen españolas morenas típicas.

Hacer clic para expandir...

Obvio, en especial las asturianas, se trajeron a miles cuando los asturianos volvieron, pero eso no quita lo otro, el noroeste concentra el mayor % de sangre bereber de toda Europa superando a Malta, que allí es más hebrea que bereber.

La pones en Colombia o en Cochabamba y se mimetiza.

En fin

Saludos Al-Celtinños.

Sicilianos???

(adjunto visible para usuarios registrados)

(adjunto visible para usuarios registrados)

No, son los irlandeses del sur.

Es que esos rasgos no salen así como así, tienen un trasfondo real.

despues de la guerra de granada los reyes catolicos repartieron moriscos por toda la peninsula puede que alli fueran mas.

Yo siempre me he preguntado qué genética tendrá esta persona gallega.

hartman4 dijo:

despues de la guerra de granada los reyes catolicos repartieron moriscos por toda la peninsula puede que alli fueran mas.

Hacer clic para expandir...

No solo eso, es anterior en varias épocas, siglo II después de cristo, siglo VIII-IX y siglo XVII, eso como mínimo, el noroeste fue el refugio de poblaciones en la península ibérica, nunca lo fue el nordeste, de ahí las enormes diferencias genéticas, la diferencia genética entre gallegos y vascos son descomunales. Es que es obvio, ¿por qué triunfo el cristianismo, una religión abrahámaica en el noroeste y no en el nordeste?

-Alexia- dijo:

Yo siempre me he preguntado qué genética tendrá esta persona gallega.

Hacer clic para expandir...

Pues que si la plantas en el Rif da el pego, ya si le pones el pañuelo, muy obvio. Tú, siendo de Valencia, seguro que no te encuentras estas caras en valencianas de origen valenciano, ¿verdad?

Topollillo dijo:

No solo eso, es anterior en varias épocas, siglo II después de cristo, siglo VIII-IX y siglo XVII, eso como mínimo, el noroeste fue el refugio de poblaciones en la península ibérica, nunca lo fue el nordeste, de ahí las enormes diferencias genéticas, la diferencia genética entre gallegos y vascos son descomunales. Es que es obvio, ¿por qué triunfo el cristianismo, una religión abrahámaica en el noroeste y no en el nordeste?

Hacer clic para expandir...

De todos modos, los ejemplos que has elegido están cogidos ex profeso, si vas a Galicia o a Asturias ni de lejos ves esos caretos morenos con asiduidad, en Asturias es bastante más fácil ver rasgos que parecen de ingleses.
Esta es una foto de una calle de Santiago:


(adjunto visible para usuarios registrados)

-Alexia- dijo:

De todos modos, los ejemplos que has elegido están cogidos ex profeso, si vas a Galicia o a Asturias ni de lejos ves esos caretos morenos con asiduidad, en Asturias es bastante más fácil ver rasgos que parecen de ingleses.
Esta es una foto de una calle de Santiago:


(adjunto visible para usuarios registrados)

Hacer clic para expandir...

Esos no parecen ingleses ni de lejos, parecen mediterráneos.

Si apred

Topollillo dijo:

Esos no parecen ingleses ni de lejos, parecen mediterráneos.

Hacer clic para expandir...

Si aprendemos a leer estaremos mejor eh...

Topollillo dijo:

Esos no parecen ingleses ni de lejos, parecen mediterráneos.

Hacer clic para expandir...

Mujeres asturianas, las dos del centro y la de arriba de la derecha podrían ser perfectamente de Leicester:

(adjunto visible para usuarios registrados)

-Alexia- dijo:

De todos modos, los ejemplos que has elegido están cogidos ex profeso, si vas a Galicia o a Asturias ni de lejos ves esos caretos morenos con asiduidad, en Asturias es bastante más fácil ver rasgos que parecen de ingleses.
Esta es una foto de una calle de Santiago:


(adjunto visible para usuarios registrados)

Hacer clic para expandir...

De esa foto saco otra conclusión: medirse por el tonalidad de piel me parece un error. Como si la piel blanca de por sí te otorgara dones especiales o diferentes. El «soy más blanco que tú» me mata.

Pues el fenómeno que observo en esa foto y esa calle es un fenómeno que voy observando por todas las calles y cada vez más, y la edad de las personas. Cada vez más las calles están pobladas de gente mayor.