Genomic reconstruction of the Pakistani Roma reveals dual South Asian ancestry, medieval bottlenecks, and the early dispersal routes of the Romani people

Abstract

The Roma people represent one of the largest and most historically enigmatic diasporas in Eurasia, illuminating human migration patterns and cultural resilience across continents. Despite extensive research on European Roma as the diaspora endpoint, the genetic legacy of their putative South Asian source populations remains critically underexplored, leaving fundamental gaps in understanding the pre-diaspora demographic structure and early dispersal dynamics. This study uniquely positions Pakistani Roma as a potential ancestral reservoir, offering a rare window into the pre-migration phase distinct from derived European Roma populations shaped by centuries of post-dispersal admixture. We analyze 82 Pakistani Roma from Punjab using high-resolution genome-wide SNP data and comprehensive mitochondrial haplogroup profiling to reconstruct their genetic origins, population structure, and historical trajectory. Analyses reveal a dual ancestry profile comprising 50–82% Indus Valley related, 20–30% Onge related, and up to 26% Steppe derived components, with three distinct subgroups exhibiting varying affinities along a South Asian to Central Western Eurasian continuum reflecting jati-like endogamy. A severe demographic bottleneck ~800 years ago coincides with medieval socio-political upheavals, while major Eurasian admixture is dated to ~660 years ago. Mitochondrial haplogroups H (45.12%) and M (26.83%) underscore dual maternal influences from West and South Eurasia. Pakistani Roma retain substantially higher South Asian ancestry than their European counterparts, establishing them as a genetically distinct population preserving the ancestral pre-diaspora state. These findings redefine the Romani origin narrative and underscore the critical value of understudied South Asian minorities in reconstructing complex human migration pathways and diaspora formation mechanisms.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s40246-026-00965-4.

Introduction

Few diasporas embody the interplay of genetic legacy, cultural endurance, and historical mystery as profoundly as the Roma. Scattered across continents yet united by shared roots, their journey from South Asia to Europe remains one of the most compelling narratives in human migration history. The Roma, often referred to as Romani or Gypsies, represent a unique ethnic group whose transnational identity and itinerant lifestyle have long fascinated anthropologists and historians. Traditionally viewed as originating from northern India and Pakistan, particularly Punjab and Kashmir [1], their history is a mosaic of migration, adaptation, and cultural resilience. Linguistic evidence anchors the Romani language within the Indo-Aryan branch of the Indo-European family, with over 60 dialects reflecting their diverse encounters and lack of a unified written tradition [2]. This linguistic diversity mirrors their anthropological complexity, suggesting a history shaped by interactions across continents. Historical narratives propose that the Roma began their westward migration between the fifth and tenth centuries, traversing the Caucasus and Anatolian Peninsula [2, 3], driven by socio-political upheavals such as the decline of the Gupta Empire and subsequent invasions [4]. By the eleventh and twelfth centuries, Ottoman expansions further propelled them into Europe, where they established a widespread presence by the fifteenth century, particularly in the Balkans, Iberian Peninsula, and East-Central Europe.

Anthropologically, the Roma’s social organization bears striking parallels to the Indian jati system, characterized by endogamy and occupational specialization [5]. This cultural continuity suggests that their migration did not erase their South Asian roots but rather adapted them to new contexts. Genetic studies have substantiated these origins, identifying Y-chromosome marker M82 (H1a) and mitochondrial haplogroups M5a1, M18, and M35b as signatures of their northern Indian ancestry [6, 7]. These genetic signatures are preserved in South Asian source populations and subsequently carried westward, rather than being defined by their presence in derived diaspora populations. The observation of Y-chromosome links to South India versus mitochondrial ties to Northwest India in European Roma highlights how derived diaspora populations may not fully represent ancestral source diversity, underscoring the need to study pre-diaspora populations like the Pakistani Roma.[8]. Genome-wide SNP analyses reveal that European Roma retain only 20–35% South Asian ancestry, reflecting substantial dilution from European gene flow from the 13th to sixteenth centuries [9, 10]. This retention level suggests that the ancestral source population likely harbored significantly higher South Asian ancestry proportions, motivating direct study of pre-diaspora populations.

Critically, previous studies have primarily characterized European Roma as the endpoint of the diaspora, whereas Pakistani Roma potentially represent a population with greater genetic continuity to the ancestral pre-migration state. Previous studies have focused on European Roma or broader South Asian populations, overlooking the anthropological and historical richness of Pakistani Roma communities [11–16]. This study bridges this gap by analyzing genome-wide SNP data from 82 Pakistani Roma individuals, positioning them as a reference for the ancestral Romani population structure. By integrating these genetic findings with anthropological and historical contexts, we establish a baseline for understanding which features of European Roma genetics reflect ancestral retention versus post-migration modification. Additionally, uniparental DNA studies provide only partial insights [17], highlighting the need for genome-wide investigations [18]. Our objectives are to (1) delineate the fine-scale genetic structure of Pakistani Roma as a potential ancestral reference population, emphasizing their cultural ties to Punjab, and (2) reconstruct the pre-diaspora demographic history of the Romani people, using European Roma data only to identify derived modifications that occurred post-migration. This approach positions Pakistani Roma not as a derivative of European Roma, but as a genetically distinct population potentially preserving ancestral diversity lost or modified in the diaspora.

Materials and methods

Ethical approval and sample collection

This study included 82 unrelated, healthy Roma individuals from Punjab Province, Pakistan. All participants were confirmed to have no biological relationship up to the third degree. Sample collection was outsourced by the Research Registry of the Office of Research, Innovation, and Commercialization (ORIC) Department, Islamia College University, Peshawar, Khyber Pakhtunkhwa, Pakistan, to a trained team of phlebotomists, ensuring standardized procedures and participant safety. The Research Registry subsequently handed over the de-identified samples to the authors, which contained only ethnic and sex information to maintain confidentiality. The study protocol received ethical approval from the Bioethics Review Committee of Islamia College University, Peshawar, Khyber Pakhtunkhwa, Pakistan (Approval No.: 866/ORIC/ICP, dated 23rd April 2023). All procedures involving human participants were carried out in strict compliance with the principles of the Declaration of Helsinki (World Medical Association, Seventh Revision, 64th WMA General Assembly, Fortaleza, Brazil, 2013) and relevant national regulations of Pakistan, and written informed consent was obtained from every participant prior to sample collection.

DNA extraction, quantification, genotyping, quality control, and merging data

We extracted genomic DNA utilizing the ReliaPrep™ Blood gDNA Miniprep System (Promega, Madison, USA), and DNA quantification was performed with the Quantifiler™ Trio DNA Quantification Kit according to the manufacturer’s instructions. Genome-wide genotyping was carried out on the Illumina WeGene Array platform (https://www.wegene.com). We aligned the raw genotype data to the human reference genome assembly GRCh37 (human_g1k_v37). We used PLINK [19] to eliminate SNPs with a missing call rate > 0.05 and individuals with a missing genotype rate > 0.1 (parameters: --geno 0.05, --mind 0.1). SNPs that failed the Hardy Weinberg equilibrium test were eliminated using PLINK’s "--hwe 0.0001" setting. Following rigorous quality control using PLINK software with parameters set to --geno 0.05, --mind 0.05, --maf 0.01, and --hwe 10⁶, third-degree related individuals were identified and removed using KING software, resulting in a final dataset of 54 individuals for downstream population genetic analysis, which encompassed 492,603 SNP loci. Additionally, we incorporated a previously published dataset of European Romani individuals to identify derived genetic signatures that distinguish post-migration diaspora populations from the ancestral South Asian source. European Roma data serve as comparative outgroup data, not as the primary reference frame. The comparative datasets included the Human Origins (HO.v03) panel (~600,000 SNPs, sourced from the Allen Ancient DNA Resource, covering modern Eurasian and South Asian populations such as Punjabi, Sindhi, Tamil and Gujarati from the 1000 Genomes Project and other public repositories) and the 1240k.v02 dataset (Reich Lab, ~1.24 million ancestry-informative SNPs, including ancient populations and IVC-related proxies like Shahr-i-Sokhta samples). The European Romani data were sourced from previously published genome-wide SNP datasets (n = 206). These datasets were selected to provide a comprehensive reference for South Asian, Central Asian, and Eurasian populations, enabling robust ancestry and admixture analyses [20]. Due to the limited number of loci after merging data from this study with the European Romani dataset and the HO/1240k datasets, we implemented a strategic merging of three datasets for comprehensive population genetic analysis: first, the data from this study (Pakistani Roma) were merged with HO.v03 and 1240k.v02 datasets, resulting in "HO_roma" and "1240k_roma", which covered 68,316 and 182,095 SNP loci, respectively; second, the European Romani data were similarly merged with HO.v03 and 1240k.v02 datasets to create "HO_euroma" and "1240k_euroma," covering 135,297 and 208,176 SNP loci; third, a combined dataset incorporating data from this study, the European Romani data, and HO/1240k was created, resulting in "HO_roma_euroma" and "1240k_roma_euroma," which covered 30,021 and 47,205 SNP loci, respectively. This comprehensive merging strategy facilitated robust analysis of genetic continuity and divergence, with Pakistani Roma serving as the primary reference for ancestral population structure and European Roma providing comparative data on derived post-migration modifications.

Analysis of population structure

We performed principal component analysis (PCA) on the "HO_roma_euroma" dataset, using the smartpca tool from the EIGENSOFT package (version 7.2.1). To project ancient individuals onto the principal components derived from modern populations, we enabled the option "lsqproject: YES." [21]. In the subsequent analysis of the "HO_roma_euroma" dataset, we removed 1,515 SNPs in strong linkage disequilibrium using PLINK1.9 [19] with the parameters "--indep-pairwise 200 25 0.4". We then applied ADMIXTURE [22] to estimate individual ancestry proportions, testing values of K (number of ancestral populations) from 2 to 10. The cross-validation error was lowest at K = 5, indicating optimal model fit with five ancestral components.

Treemix analysis

We utilized TreeMix [23] on the "HO_roma_euroma" dataset to build a rooted maximum-likelihood tree, with Mbuti designated as the root for the populations in our study. This method was employed to infer phylogenetic relationships. The results were visualized using plotting_funcs.R script.

Pairwise Fst genetic distance

To estimate pairwise Fst genetic distance, we used smartpca from the EIGENSOFT package (version 7.2.1) [21] on the "HO_roma_euroma" dataset, with default parameters along with the options ‘inbreed: YES’ (to account for the high level of endogamy and potential inbreeding in the Roma population, which can affect Fst estimation) and ‘fsonly: YES’ (to output Fst values only)b. Subsequently, we used the "fitch" program in the PHYLIP package [24] for constructing phylogenetic trees (https://phylipweb.github.io/phylip/phylip.html).

Analysis based on shared IBD segments

Identity-by-descent (IBD) fragments defined as chromosomal segments inherited from a common ancestor without recombination were detected between each pair of individuals using Refined IBD (version 12Jul18) [25] on the "HO_roma" dataset. Adjacent IBD segments were merged using merge-ibd-segments (version 12Jul18) [26]. For pairwise population comparisons, we normalized the total shared IBD by dividing by the product of the sample sizes. Effective population size (Ne) changes over time were estimated using IBDNe (version v23Apr20) [25], assuming a generation time of 29 years.

f-statistical analysis

We calculated f3 and f4 statistics using the qp3Pop and qpDstat programs (version 950) within the ADMIXTOOLS package [27] on the "HO_roma" and "1240k_roma" datasets. For the outgroup-f3 analyses, we used the format f3 (Target, X; Mbuti), to evaluate the genetic affinity between the Roma and various modern populations from Eurasia continent. Populations with insufficient SNP overlap (less than 30,000) were excluded from the analysis. Additionally, we computed f4 statistics with the parameter ‘f4mode: YES’, to formally assess whether other referred populations shared more alleles with our target population, based on significant positive or negative f4 values.

Pairwise qpWave and qpAdm-based admixture modeling

We applied pairwise qpWave, as implemented in the ADMIXTOOLS package, on the ‘1240k_roma’ dataset to determine whether selected population pairs were genetically homogeneous with respect to a set of outgroups (Mbuti.DG, Tianyuan, DevilsCave_N.SG, YR_MN, Longlin, Kostenki14, Russia_EHG, Tyumen_HG, Luxembourg_Loschbour, Ganj_Dareh_N, Anatolia_N). A p value >0.05 for rank = 0 was interpreted as evidence of genetic homogeneity between the population pairs relative to the outgroups, while a p value <0.05 for rank = 0 suggested that at least two distinct ancestral lineages were required to explain the genetic relationships between the population pairs and the outgroups. We used the qpAdm from the ADMIXTOOLS package [27], with its default settings, to estimate the admixture proportions of a target population. This was done by analyzing shared genetic drift with a predefined set of outgroups (Mbuti.DG, Tianyuan, DevilsCave_N.SG, YR_MN, Longlin, Kostenki14, Russia_EHG, Tyumen_HG, Luxembourg_Loschbour, Ganj_Dareh_N).

Estimating the admixture time with ALDER

We employed ALDER (version 1.03) on the "HO_roma" dataset to infer the timing of admixture events, using Punjabi and Georgian as the two reference populations. For this analysis, we selected the parameter ‘checkmap: NO’. ALDER, which calculates a weighted linkage disequilibrium (LD) statistic, is effective in both estimating admixture proportions and dating these events [28].

mtDNA analysis

We genotyped the lineage-informative SNPs (LISNPs) in mitochondrial DNA using the Illumina array and an in-house script adhering to the guidelines provided by the International Society of Genetic Genealogy (ISOGG; http://www.isogg.org/) and mtDNA Tree Build 17 [29]. We utilized Haplogrep2 for mitochondrial haplogroup classification [30]. The Illumina WeGene Array targeted ~600,000 genome-wide SNPs, yielding 492,603 post-QC loci. Mitochondrial DNA genotyping included the full control region (HVR-I/II) and diagnostic coding SNPs, classified via HaploGrep2 [30].

Results

Preliminary analysis

Mitochondrial haplogroup diversity in the Pakistani Roma

Mitochondrial DNA (mtDNA) analysis of 82 Roma individuals from Punjab, Pakistan, reveals a distinctive maternal genetic profile that reflects both deep South Asian roots and layers of historical admixture during westward migrations (Supplementary Fig. 1, Supplementary Table 1). The most prevalent haplogroup is H (45.12%), a lineage common across Europe, West Asia, and parts of Central Asia [9, 31]. The second most frequent lineage is haplogroup M (26.83%), a macro-haplogroup that originated in South Asia over 50,000 years ago and is considered a hallmark of indigenous maternal ancestry in the Indian subcontinent [32, 33]. Haplogroup U (12.20%) particularly subclades like U7, commonly found in South Asia, and U5/U3, more typical of West Eurasia. Minor maternal lineages include haplogroups T (4.88%), K (2.44%), and F, X, and N (1.22% each) which are typically associated with Neolithic expansions, Bronze Age movements, and later historical contacts across West Eurasia [34]. The mtDNA composition indicates a combination of South Asian maternal ancestry (primarily haplogroup M) with West and Central Eurasian components (haplogroups H, U, T, and K). The high frequency of haplogroup H in this population contrasts with its rarity in most indigenous South Asian groups, suggesting historical contact with West Eurasian populations [9, 31]. Haplogroup U subclades show affiliations with both South Central Asia and regions extending to Iran, the Caucasus, and Eastern Europe [35, 36]. Together, these patterns indicate a maternal gene pool structure characterized by geographic dualism, incorporating both South Asian and West Eurasian lineages.

Effective population size analysis of the Pakistani Roma population

Effective population size (Ne) analysis of the Pakistani Roma, using the IBDNe method (version v23Apr20) [25] with a 29-year generation time (Fig. 1). The analysis reveals three demographic phases. Prior to the bottleneck, Ne remained stable at approximately log values ~104. Approximately 800 years ago, Ne declined sharply to near 102. Following this minimum, Ne exhibits a gradual increasing trajectory indicating post-bottleneck expansion within the Punjab region. Before the bottleneck, a stable and relatively high Ne (log values ~10⁴) reflects the ancestral demographic state in the Punjab region, indicative of a large, sedentary population with consistent growth and genetic diversity. Approximately 800 years ago, a severe bottleneck occurred, with Ne dropping to near 10², suggesting a near population collapse. This bottleneck is temporally distinct from demographic events observed in European Roma populations, which exhibit separate founder effects (see Discussion). The gradual post-bottleneck recovery evident in increasing Ne values reflects population expansion and admixture within the Punjab region. Importantly, this bottleneck signature is distinct from the separate founder effects experienced by European Roma populations during their westward dispersal; Pakistani Roma and European Roma underwent independent demographic collapses driven by different historical pressures in different geographic locations. These findings provide insights into the demographic history of the ancestral Romani reservoir population in Punjab, the localized pressures shaping its genetic diversity, and the independent trajectories of populations that remained in South Asia versus those that eventually dispersed toward Europe.

Fig. 1.

Demographic history of Pakistani Roma. Effective population size (Ne) trajectory inferred using IBDNe (29-year generation time). The severe bottleneck ~800 years ago (Ne drop to ~10²) predates the ~660-year-ago admixture event and is temporally distinct from European Roma founder effects

Homogeneity testing between individuals

Pairwise qpWave analysis, visualized in a heatmap, assessed genetic relationships within the Pakistani Roma population using diverse outgroups. This grouped individuals into three primary clusters Roma_Pakistan1, Roma_Pakistan2, and Roma_Pakistan3 indicating distinct genetic substructures (Supplementary Fig. 2). The heatmap shows pairwise genetic affinities, with lighter shades indicating weaker and darker shades stronger similarities. Most individuals cluster consistently within these three groups, demonstrating clear differentiation. However, individual 1UD21210100037 exhibited strong genetic heterogeneity across all three groups, identifying them as an outlier potentially due to a unique genetic history, recent admixture, or sampling error. To maintain clarity of genetic distinctions, this outlier was labeled Roma_Pakistan_o and excluded from subsequent population-level analyses. Focusing on the three cohesive groups enhances the resolution of demographic and genetic investigations into the Pakistani Roma population, providing a basis for further analyses of admixture, population dynamics, and historical migrations.

Initial qualitative analysis

Overview of genetic affinity and population structure

Analysis of genome-wide SNP data from 54 unrelated Pakistani Roma individuals provides insights into their genetic relationships and population structure, examining three distinct groups: Roma_Pakistan1, Roma_Pakistan2, and Roma_Pakistan3. These groups represent substructure within the ancestral source population, providing the baseline against which derived European Roma diversity can be compared. This aims to unravel the Pakistani Roma’s genetic history within a broader regional and global context, shedding light on ancestral composition, historical admixture, and genetic affinities with modern South Asian and Eurasian populations.

Principal Component Analysis (PCA) was performed at two scales to assess population structure. Broad continental PCA including global reference populations (African, East Asian, Siberian, Oceanian, European, South Asian, and West Asian groups) reveals that Pakistani Roma cluster within the South Asian genetic continuum, with European Roma positioned distinctly (Fig. 2A). However, highly divergent outliers compress South Asian variation along a single axis, limiting resolution of internal structure. To resolve fine-scale affinities, a focused PCA was computed excluding distant outliers (African, East Asian, Siberian, and Oceanian populations) (Fig. 2B). In this analysis, the three Pakistani Roma subgroups exhibit distinct positions along a North–South cline. Roma_Pakistan1 aligns toward modern southern South Asian populations. Roma_Pakistan3 shifts toward modern northern South Asian and ancient Steppe/Central Asian populations. Roma_Pakistan2 occupies an intermediate position. European Roma cluster separately from all Pakistani Roma subgroups [9, 37–39].

Fig. 2.

Population structure in global and regional contexts. A Broad continental PCA including African, East Asian, Siberian, Oceanian, European, and South Asian populations (PC1: 5.5% variance, PC2: 0.7% variance). Outlier populations compress South Asian variation along PC1. B Focused PCA excluding distant outliers (PC1: 1.18% variance, PC2: 0.26% variance) reveals fine-scale structure: Pakistani Roma subgroups align along a North-South Indian cline (Roma_Pakistan1 toward southern South Asians; Roma_Pakistan3 toward northern South Asian/Steppe populations; Roma_Pakistan2 intermediate). European Roma cluster separately. Ancient Indus Valley Civilization (IVC) samples shown in gray

ADMIXTURE analysis further elucidates genetic structure and ancestry components (Supplementary Fig. 3). The Pakistani Roma predominantly exhibit a South Asian genetic signal, consistent with their origins and ties to regional populations like Punjabis, Sindhis, and Makranis. This reflects shared regional heritage. They also display genetic contributions from Central-Western Eurasia, highlighting interactions during migration. The blue component, subsequently prominent in European Roma, signifies shared ancestry originating in the South Asian source population prior to European dispersal, while green and yellow components indicate ancestral gene flow from Iranian farmer and Steppe-related sources that contributed to the reservoir population before westward migration. Roma_Pakistan3 shows a stronger representation of these components, indicating proximity to Central and Western Eurasian migration corridors. Compared to European Roma, Pakistani Roma exhibit less European admixture, reflecting geographic separation and localized gene flow.

qpWave analysis (Fig. 3) highlights ancestral genetic diversity among Pakistani Roma subgroups, representing pre-diaspora population structure that was subsequently carried, in modified form, into European Roma populations. Roma_Pakistan1 and Roma_Pakistan3 align closely with Punjabis and Sindhis, respectively, reflecting regional interactions and trade networks. qpAdm results (Fig. 4) reveal substantial ancestry from the Indus Valley Civilization (IVC), with varying degrees of Steppe-related ancestry. Roma_Pakistan3 shows a pronounced IVC component, suggesting a strong genetic link to ancient South Asian populations. This finding aligns with the Roma’s historical trajectory, tracing their roots to the culturally rich IVC region and subsequent migrations shaped by socio-political changes, including the decline of the Gupta Empire [6].

Fig. 3.

Genetic homogeneity among Pakistani Roma subgroups. Heatmap of p-values from pairwise qpWave tests (rank = 0) assessing whether population pairs are genetically homogeneous relative to outgroups. Blue indicates significant heterogeneity (p < 0.05); light shades indicate homogeneity. Roma_Pakistan1 shows homogeneity with Punjabi; Roma_Pakistan3 with Sindhi

Fig. 4.

Ancestry composition of Pakistani and European Roma. qpAdm modeling of three-way admixture (IVC-related, Onge-related, Steppe-related) with outgroup set (Mbuti.DG, Tianyuan, DevilsCave_N.SG, YR_MN, Longlin, Kostenki14, Russia_EHG, Tyumen_HG, Luxembourg_Loschbour, Ganj_Dareh_N). Pakistani Roma retains 48–82% IVC ancestry, contrasting with European Roma (predominantly Steppe ancestry)

Collectively, these analyses present a coherent picture of the Pakistani Roma’s genetic history, shaped by South Asian origins, westward migrations, and endogamy. They underscore the complex interplay of ancient ancestry, migration, and regional interactions, offering insights into the broader historical dynamics of Roma populations. The complementary methodologies provide a comprehensive framework for understanding their genetic structure and historical demographics, highlighting their enduring South Asian connection and intricate migratory history.

Genetic insights into the migration, ancestry, and historical admixture events of the Pakistani Roma population

The Neighbor-Joining tree based on pairwise Fst values reveals clustering patterns among Pakistani Roma and comparative populations (Supplementary Fig. 4, Supplementary Table 2). Pakistani Roma cluster closely with South Asian populations, particularly those from northwestern India and Pakistan. European Roma forms a distinct cluster adjacent to indigenous European populations, separated by long branch lengths from South Asian groups. The topology positions Pakistani Roma between Central-South Asian and European clusters, with shorter genetic distances to Punjabi, Sindhi, and Kalash populations compared to European groups [39].

Maximum-likelihood phylogenetic analysis using TreeMix (Fig. 5) shows Pakistani Roma clustering with northern South Asian populations (Punjabi, Sindhi, Balochi, Brahui), distinct from the European Roma cluster. European Roma populations (Hungary, Slovakia, Ukraine, Lithuania, Estonia, Serbia, Croatia, Bulgaria, Romania, Greece, Spain, Portugal, Wales) occupy a peripheral branch with extended drift lengths. No migration edges were detected between Pakistani Roma and European Roma lineages. The tree topology indicates substantial genetic drift in the European Roma lineage compared to Pakistani Roma, with the latter positioned within the South Asian genetic continuum.

Fig. 5.

Phylogenetic relationships among Pakistani Roma, European Roma, and reference populations. Maximum-likelihood tree inferred using TreeMix (rooted with Mbuti). Branch lengths represent genetic drift. Pakistani Roma cluster with northern South Asian populations; European Roma show long branch lengths indicating extensive drift. No migration edges detected between Pakistani and European Roma lineages

IBD analysis of segments >1 cM reveals shared haplotypes between Pakistani Roma and regional populations (Fig. 6, Supplementary Table 3). The highest IBD sharing occurs with Punjabi and Sindhi populations, followed by Makrani and Kalash groups. Normalized IBD scores indicate substantial haplotype sharing between the three Pakistani Roma subgroups and northern South Asian populations. Shared segment lengths and frequencies suggest recent common ancestry between Pakistani Roma and these groups within the past 50 generations. European Roma show minimal IBD sharing with Pakistani Roma, consistent with long-term reproductive isolation.

Fig. 6.

Identity-by-descent sharing patterns. Normalized IBD sharing between Pakistani Roma subgroups and reference populations. A Total IBD scores; B Mean segment lengths. Higher sharing with Punjabi and Sindhi populations supports regional ancestry within Punjab

NJ tree, TreeMix, and IBD analyses consistently demonstrate close genetic affinity between Pakistani Roma and northern South Asian populations, with measurable genetic distance between Pakistani and European Roma populations.

Advanced qualitative analysis

Qualitative analysis using f-statistics provides insights into the Pakistani Roma’s genetic ancestry and admixture history.

Outgroup f3-statistics, using the "HO_roma" dataset, measure shared genetic drift between Pakistani Roma and various modern populations relative to Mbuti (Fig. 7). Results show the strongest shared genetic drift with South Asian populations, particularly Punjabis, Sindhis, and Gujaratis (Supplementary Table 4). Punjabi populations exhibit the highest f3 values, signifying the closest genetic affinity and suggesting an origin in the Punjab region. High f3 values with Sindhis and Gujaratis imply genetic continuity or historical interactions.

Fig. 7.

Genetic affinity of Pakistani Roma to Eurasian populations. Outgroup f3-statistics (Target, X; Mbuti) showing shared genetic drift. Higher values indicate closer affinity. Pakistani Roma exhibit strongest affinity to northern South Asian populations (Punjabi, Sindhi, Gujarati)

F4-statistics, analyzed as f4 (Mbuti, X; Roma, modern South Asia), complement f3 results by characterizing ancestral admixture events that occurred in the South Asian source population prior to major westward dispersal. Roma_Pakistan3 exhibits the strongest Steppe-related ancestry, while Roma_Pakistan1 shows the least (Fig. 8, Supplementary Table 5A&B).

Fig. 8.

Steppe-related ancestry gradients. f4-statistics (Mbuti, Steppe_reference; Roma_Pakistan, X) revealing differential Steppe ancestry across subgroups. Roma_Pakistan3 shows highest Steppe-related ancestry; Roma_Pakistan1 shows least, consistent with geographic substructure in the ancestral range

The combined f3 and f4 results provide a robust genetic narrative, emphasizing a Punjabi-Sindhi origin with admixture from Central and Western Eurasian populations, reflecting ancestral ties, historical migrations, and interactions, offering insights into their population structure and evolutionary journey.

Quantitative analysis

qpWave analysis

Pairwise qpWave analysis tested whether Pakistani Roma subgroups and reference populations form genetically homogeneous clades relative to outgroups (Mbuti.DG, Tianyuan, DevilsCave_N.SG, YR_MN, Longlin, Kostenki14, Russia_EHG, Tyumen_HG, Luxembourg_Loschbour, Ganj_Dareh_N, Anatolia_N) (Fig. 3, Supplementary Table 6). Roma_Pakistan1 exhibited genetic homogeneity with Punjabi (chi-square = 22.575, p = 0.0124). Roma_Pakistan3 showed genetic homogeneity with Sindhi (p = 0.375). Roma_Pakistan2 did not demonstrate significant homogeneity (p < 0.05) with any tested reference populations, indicating rejection of the null hypothesis of single-component ancestry for this subgroup.

qpAdm ancestry estimation

Admixture modeling using qpAdm quantified ancestral contributions for Pakistani Roma subgroups using the same outgroup set (Fig. 4, Supplementary Table 7). Ancestry proportion estimates indicate that Roma_Pakistan1 comprises approximately 50% Indus Valley Civilization (IVC)-related ancestry (proxied by Shahr-i-Sokhta_BA1), 30% Onge-related ancestry, and 20% Steppe-related ancestry (proxied by Russia_Alan.SG). Roma_Pakistan2 shows a composition of approximately 48% IVC-related, 26% Onge-related, and 26% Steppe-related ancestry. Roma_Pakistan3 exhibits a predominance of IVC-related ancestry at approximately 82%, with 18% Onge-related ancestry and minimal Steppe-related contribution. For comparative reference, European Roma populations (n = 206) exhibited ancestry estimates of approximately 79.8–85.6% Steppe-related and 10.7–20.2% Onge-related ancestry (Supplementary Table 8).

Admixture dating

ALDER analysis estimated the timing of population admixture using Punjabi and Georgian as reference populations with a generation time of 29 years (Supplementary Fig. 5, Supplementary Table 9). The analysis estimated admixture occurred approximately 660 years ago (exponential decay parameters ~29). Analysis parameters included ‘checkmap: NO’ due to merged dataset constraints.

Discussion

The genetic analyses presented above provide a quantitative foundation for reconstructing the population history of Pakistani Roma. Here, we interpret these statistical patterns within their anthropological and historical contexts, specifically contrasting the Pakistani Roma profile with that of European Roma to distinguish ancestral source characteristics from derived diaspora modifications. We emphasize that Pakistani Roma represent a population with substantial genetic continuity to the pre-diaspora ancestral state, whereas European Roma constitute a derived population whose genetic profile has been substantially modified by 800+ years of independent demographic history. Outgroup-f3 statistics affirm the deep-rooted affinity of Pakistani Roma with northern South Asian populations, particularly Punjabis and Sindhis, establishing the Punjab-Sindh region as the demographic reservoir from which Romani diaspora populations ultimately derived [9, 13]. This genetic proximity underscores a historical continuity with the region’s inhabitants, whose social structures, including the jati-like endogamy observed among the Roma, likely influenced their community organization even as they migrated [5, 15]. The retention of such systems reflects the resilience of cultural practices in preserving identity amidst centuries of movement and adaptation.

The prominence of mitochondrial haplogroup M (26.83%) reflects maternal lineages tied to South Asia’s ancient populations, while haplogroup H (45.12%) suggests maternal contributions acquired during westward movements [29, 37]. The high frequency of haplogroup H (45.12%) rare in indigenous South Asian populations but subsequently dominant in European Roma indicates that this West Eurasian maternal component was present in the ancestral Punjabi source population prior to major dispersal, likely reflecting historical contact along northwestern trade corridors, rather than being acquired by Pakistani Roma through recent European gene flow. This establishes Pakistani Roma as preserving pre-diaspora maternal diversity that was subsequently carried westward and retained in derived populations through founder effects. These findings highlight the Roma’s role as carriers of cultural and genetic legacies across continents, embodying a synthesis of South Asian heritage and Eurasian encounters [37, 38]. F4 statistics reveal a significant Steppe and Central Eurasian ancestry in the ancestral reservoir population, particularly pronounced in Roma_Pakistan3, aligning with historical migrations of pastoralist groups into South Asia around 2000 BCE. These ancestral components were subsequently inherited by westward-migrating populations, rather than being acquired during the diaspora itself [6, 27, 40]. This admixture event, dated to at least 660 years ago via ALDER, occurred in the ancestral reservoir population and coincides with medieval disruptions such as Mongol invasions and shifting power dynamics during the Delhi Sultanate. These events likely triggered or accompanied the initial displacement from Punjab that eventually led to the westward dispersal toward Europe [28, 41]. The demographic trajectory reveals two critical temporal signals: a severe bottleneck ~800 years ago marking population collapse in medieval Punjab, followed by detectable Eurasian admixture ~660 years ago (Fig. 1, Supplementary Fig. 5). We posit that these sequential events demographic pressure followed by population restructuring precipitated the initial displacement from the ancestral range that eventually culminated in the westward dispersal toward Europe, while the reservoir population that remained in Punjab underwent independent recovery.

Specifically, this bottleneck evidenced by a sharp decline in effective population size to Ne ~ 10² (Fig. 1) reflects localized demographic pressures in medieval Punjab [29, 42]. We emphasize that this collapse represents a localized event in medieval Punjab, not a migration-associated founder effect. This population remained in South Asia and underwent independent recovery, while European Roma represent a parallel, derived lineage that experienced separate, additional bottlenecks during westward dispersal. We explicitly reject any model implying back-migration from Europe; the temporal coincidence of bottlenecks reflects parallel responses to broader medieval Eurasian turmoil rather than shared demographic history. The demographic collapse ~800 years ago corresponds to the early to mid thirteenth century CE (1200–1250 CE), aligning with documented socio-political upheavals in medieval Punjab. This period witnessed the decline of the Ghurid dynasty following the death of Muhammad of Ghor in 1206 CE, and critically, the first Mongol invasion of South Asia (1221–1222 CE) when Genghis Khan’s forces pursued Jalal ad-Din Mingburnu into the Punjab region. Historical records specifically document those Mongol forces besieged Multan for 40 days in 1222 CE, creating widespread regional disruption [43, 44]. Additionally, paleoclimate evidence indicates deteriorating conditions in the Indus valley system during this period, potentially contributing to agricultural stress [45]. These converging factors dynastic transition, foreign military incursion, and possible environmental deterioration provide a coherent historical framework for the severe genetic bottleneck detected.

Comparative population genetics studies indicate that severe bottlenecks ~600–1000 years ago characterize several populations subjected to medieval socio-political upheavals. Notably, Ashkenazi Jewish populations exhibit a founder bottleneck estimated to have occurred 25–32 generations ago (~600–800 years ago), reducing the population to approximately 350 individuals, associated with persecution and geographic isolation in medieval Europe [46]. Similarly, Siberian populations (such as the Yakut) show evidence of bottlenecks ~800–1000 years ago coinciding with the expansion of the Mongol Empire [47]. Medieval European populations including Albanians also exhibit demographic fluctuations consistent with bottlenecks during the 500–1000 CE period [48]. These parallel cases demonstrate that medieval political disruptions particularly those involving the Mongol Empire generated comparable demographic signatures across diverse Eurasian populations, supporting the interpretation that the Pakistani Roma bottleneck reflects localized responses to these broad historical forces rather than a migration-specific founder event.

This demographic collapse likely resulted from socio-political upheavals including dynastic decline, Mongol incursions, and regional conflicts, rather than westward migration. The subsequent recovery reflects population resilience through endogamy and regional expansion within South Asia [15, 49].

The dual PCA strategy illustrates how analytical scale affects the inference of population structure. While the broad continental analysis (Fig. 2A) establishes the general South Asian affinity of Pakistani Roma, the compression of variation by highly divergent global outliers obscures critical internal differentiation. The focused South Asian PCA (Fig. 2B) resolves this limitation, revealing that the three subgroups occupy distinct positions along a North South Indian cline. This pattern demonstrates genuine ancestry gradients within the ancestral Punjabi source population ranging from southern South Asian affinity (Roma_Pakistan1) to Steppe-influenced northern profiles (Roma_Pakistan3) rather than recent admixture artifacts or sampling noise. The clear separation of European Roma in both contexts reinforces their status as a derived lineage that underwent extensive post-migration drift, whereas Pakistani Roma retain structured diversity characteristic of the pre-diaspora reservoir. These substructure patterns corroborate the qpAdm and IBD findings, collectively supporting a model of jati-like endogamy maintaining distinct ancestral components within the Punjab region prior to the major westward dispersal. The three distinct Pakistani Roma subgroups Roma_Pakistan1, Roma_Pakistan2, and Roma_Pakistan3 illustrate substructure within the ancestral source population prior to any major dispersal (Fig. 4). Whereas European Roma exhibit predominantly Steppe-related ancestry (79.8–85.6%) reflecting extensive post-migration admixture [9, 39], the Pakistani Roma subgroups retain markedly higher Indus Valley Civilization-related ancestry (48–82%). Roma_Pakistan1’s proximity to southern South Asians suggests stronger retention of indigenous Punjab ancestry [13], while Roma_Pakistan3’s alignment with Central Western Eurasian populations points to greater pre-diaspora exposure to trade corridors such as the Silk Road [27, 40]. This geographical pattern corresponds to their genetic profiles: Roma_Pakistan3 exhibits pronounced IVC affinity (~82%) with minimal Steppe contribution, reflecting deeper indigenous ancestry, while Roma_Pakistan2’s elevated Steppe component (~26%) and intermediate position may reflect unique isolation or selective intermingling within the ancestral range [40]. These distinct profiles illustrate the range of ancestral diversity present within the pre-diaspora source population (Fig. 4, Supplementary Table 7). The phylogenetic analyses (Fig. 5) corroborate this distinction: the absence of migration edges between Pakistani and European Roma lineages, combined with the extensive drift evidenced by long branch lengths in the European Roma cluster, confirms that these populations share distant common ancestry from a South Asian source rather than recent genetic continuity. The IBD patterns (Fig. 6) similarly reflect this substantial haplotype sharing between Pakistani Roma and northern South Asian groups contrasts with minimal sharing between Pakistani and European Roma, consistent with independent demographic histories following initial population divergence [6, 23, 39]. This divergence highlights distinct demographic histories: Pakistani Roma maintained substantial genetic continuity with South Asian origins as a pre-diaspora reservoir population, while European Roma experienced substantial genetic modification as a derived diaspora community [38, 39].

Anthropologically, these genetic patterns resonate with the Roma’s itinerant lifestyle and social cohesion. Their retention of South Asian ancestry, quantified at 50–82% IVC-related in qpAdm models, mirrors their preservation of cultural practices, such as oral traditions and kinship-based organization, despite centuries of migration [9, 10]. The Steppe ancestry (up to 26% in Roma_Pakistan2) and Onge contributions (20–30%) evoke a deeper historical narrative, connecting the Roma to ancient population movements across the Indian subcontinent and beyond, while their Eurasian admixture reflects the dynamic interplay of conquest, trade, and diaspora that shaped medieval South Asia and Central Asia [27, 40].

These findings position Pakistani Roma as a critical reference for understanding the Romani narrative, embodying the ancestral genetic and cultural state from which diaspora populations derived, including subsequent Eurasian historical encounters. Their genetic diversity, framed by anthropological resilience and historical migrations, calls for further interdisciplinary research combining high-coverage genomics, archaeological evidence, and ethnographic studies to fully unravel the timelines and pathways of their extraordinary journey [2, 37].

Conclusion

This study aimed to reconstruct the genetic origins, migration history, and population structure of the Pakistani Roma as a potential ancestral reference population, establishing the pre-diaspora baseline against which derived European Roma populations can be understood. Our work bridges anthropology, history, and genomics to illuminate the pre-diaspora phase of the global Roma narrative.

Our analyses reveal that the Pakistani Roma harbor 50–82% ancestry related to the Indus Valley Civilization, with additional contributions from Onge-related (20–30%) and Steppe pastoralist (up to 26%) sources. These proportions likely represent the ancestral state prior to westward dispersal, subsequently modified in European Roma through additional bottlenecks and admixture. Mitochondrial haplogroups M (26.83%) and H (45.12%) reflect dual maternal origins in South Asia and West Eurasia, respectively, consistent with a westward migration beginning ~800 years ago. A severe demographic bottleneck ~800 years ago, coinciding with medieval upheavals including the first Mongol invasions of Punjab, likely precipitated the subsequent founder driven exodus. Eurasian admixture detected ~660 years ago aligns with the onset of the European dispersal. Furthermore, three genetically distinct subgroups (Roma_Pakistan1–3) form a cline from southern South Asian to Central Western Eurasian affinity, revealed through fine scale genetic analysis, and shaped by endogamy and differential admixture.

These findings redefine the Roma’s origin story by positioning the Pakistani Roma as a vital reservoir of ancestral Romani genetic and cultural continuity, with European Roma representing a derived population that experienced substantial post-migration modification. The study resolves longstanding ambiguities by demonstrating that South Asian roots particularly in the Punjab-Sindh corridor are genetically robust and quantifiable, while also revealing how historical migrations layered Eurasian ancestry onto this foundation. Practically, this work provides a reference framework for forensic, anthropological, and public health initiatives involving underrepresented South Asian minority groups.

While our sample size (n = 82 for mtDNA; n = 54 for genome-wide data) offers high-resolution insights, it represents a geographically focused cohort from Punjab and may not capture the full diversity of Roma communities across Pakistan. Additionally, admixture dating using ALDER may be affected by limited SNP overlap in merged datasets, warranting validation with higher-coverage sequencing in future studies.

Future work should expand sampling to Roma communities in Sindh, Balochistan, and Kashmir to test regional heterogeneity and validate the substructure patterns identified in the Punjab cohort. Ancient DNA from Indus Valley and early medieval South Asian contexts could directly anchor ancestry models and clarify the specific historical events associated with the ~800-year demographic collapse.

As guardians of a dual heritage rooted in the ancient soils of the Indus Valley yet shaped by millennia of movement, the Pakistani Roma embody the resilience of diasporic identity. This study not only reaffirms their centrality to the global Romani origin story but also underscores the power of interdisciplinary genomics to recover lost chapters of human history, offering a model for studying other historically mobile and marginalized populations worldwide.

Author contributions

Atif Adnan (AA) contributed to the study design, data analysis, and manuscript preparation. Haodong Chen (HC) contributed to data analysis. Supervision, study design, and funding were provided by Chuan-Chao Wang (CCW). Critical feedback was provided by Muhammad Ilyas (MI) and Allah Rakha (AR).

Funding

Not applicable.

Availability of data and materials

The PLINK-formatted genotype dataset is available to interested researchers upon reasonable request. Requests should be directed to the corresponding authors, (AA and CCW). The data will be provided in accordance with ethical and legal requirements governing the sharing of genetic information.

Declarations

Ethics approval and consent to participate

This study was reviewed and approved by the Ethics Committee of Islamia College Peshawar, Khyber Pakhtunkhwa, Pakistan (Approval No. 866/ORIC/ICP, dated April 3, 2023). Written informed consent was obtained from all participants prior to their inclusion in the study. All experimental procedures were conducted in accordance with the ethical principles outlined in the Declaration of Helsinki (1964), as amended.

Consent for publication

All participants provided informed consent for the publication of aggregated and anonymized genetic data derived from their samples, with all direct identifiers removed to protect privacy.

Competing interests

The authors declare no competing interests.