Admixture and DATES analysis of UP-Bihar Syeds

Applying multiple methods for crowd-sourced data analysis

Recent Findings and the FTDNA Banu Hashim Project

Recent findings from the FTDNA Banu Hashim haplogroup project strongly hint that J-L859 is the best contender haplogroup for direct patrilineal relationship with Banu Hashim tribe. We identified several South Asian samples within this haplogroup tree. By examining the TMRCA of the South Asian J-L859 samples, we estimate that some of these migrations likely occurred around 800 years ago.

**TMRCA for haplogroup J-Y225178 on yFull

In this blog, our focus will be on the autosomal impact of these migrations on South Asia. We collected samples from Syeds living in regions close to where these J-L859 haplogroups were identified, with the goal of determining whether any traceable West Asian genetic impact, possibly resulting from historical intermarriage, could be detected.

Big disclaimer: admixture can’t prove someone to be a Syed. It depends more on Haplogroup or your family history

Our crowdsourced Dataset

For this study, we gathered 21 crowd-sourced samples from Uttar Pradesh (UP) and Bihar in the Central Plains region of India. The raw data originated from one of four sources: 23andMe, Ancestry, FTDNA, or WGS. Approximately 25% of these crowd sourced samples tested positive for the J-L859 mutation, though there is some sample bias as these participants were more willing to share their data.

Most UP Syed samples were Shia, while nearly all Bihari Syed samples were Sunni, primarily from South Bihar. We will keep the data anonymous and refrain from disclosing any identifying information.

Figure 0: Map showing South Bihar region

Figure 0: Map showing UP (Uttar Pradesh) region

Methods

To ensure the robustness of our findings, we triangulated our analysis using several methods: 1) qpAdm group runs, 2) qpAdm individual models, 3) DATE admixture model, 4) PCA plots, 5) Y-haplogroup distribution and 6) commercial G25 models. If the migration story is traceable, we believe all these different models should converge, corroborating one another in a nice manner.

qpAdm Group Runs

For the autosomal admixture analysis, we used the qpAdm tool developed by Reich Lab at Harvard, a frequentist admixture tool extensively used by various research labs (e.g., Harney et al., "Assessing the performance of qpAdm") and used in published studies (e.g., Narasimhan et al The Formation of Human Populations in South and Central Asia).

A qpAdm model requires two sets of populations: "right" populations (reference populations) and "left" populations (those used to build the model). We chose right populations commonly used in studies of South Asian ancestry. Many of the left and right populations are available in the Reich Lab’s 1240k dataset.

Most South Asians, particularly Indians, can be modeled using three primary components: 1) South Asian Hunter-Gatherers (also referred to as AASI), the region’s ancient inhabitants, 2) Indus N Ancestry, which mixed with the Hunter-Gatherers 5,000–7,000 years ago, contributing to the Indus Valley Civilization, and 3) Steppe Hunter-Gatherers, who brought Indo-European languages to South Asia around 3,500 years ago (1500 BCE to 500 BCE).

Our goal with admixture model, was to assess whether Syeds of Indian origin possess a distinct fourth component indicative of West Asian ancestry. We tested this hypothesis using several proxy populations from West Asia, including Lebanon_Medieval.SG, Iran_Hasanlu_IA, and Syria_Tell_Qarassa_Umayyad.

To ensure accuracy, we monitored the total number of SNPs and sample sizes, as low SNP counts or small sample sizes can produce misleading results (Harney et al).

Figures 1 and 2 illustrate the qpAdm group runs using 5 and 6 samples for UP and Bihari Syeds, respectively. The valid SNP counts were 640k and 417k, which provide strong coverage from the Reich dataset.

We observed detectable amounts of West Asian ancestry, using Iran_Hasanlu_IA as proxy sample. The model passed successfully with both groups with about 15% West Asian admixture. (p value 0.05 and low std errors).

Figure 1: UP Syed group runs using Iran_Hasanlu_IA source

Figure 2: Bihar Syed group runs using Iran_Hasanlu_IA source

We also did many individual sample runs with different West Asian samples ( Syria Tell Qarassa, Lebanon Medieval and Iran_Hasanlu_IA). The results in figure 3 & 5, show that most of these samples have on an average of 5-6 % detectable Syrian Tell Qarassa type ancestry among others.

Figure 3: Testing individual UP Syed samples with Syria Tell Qarassa sample.

Figure 4: Testing individual UP Syed samples with Lebanon Medieval and Iran_Hasanlu_IA samples

Figure 5: Testing individual Bihar Syed samples with Syria Tell Qarassa sample

Figure 6: Testing individual Bihar Syed samples with Lebanon Medieval and Iran_Hasanlu_IA sources

Date of West Asian admixture

We aimed to determine when Syed migrants from West Asia intermarried with local South Asian populations. To do this, we utilized the DATES software tool developed by Moorjani Lab, a tool widely used and validated in various research publications.

Since almost all South Asians have Iran Farmer-type ancestry dating back 5,000–7,000 years, it was crucial for us to ensure that the West Asian admixture dates for Syeds fell within the Medieval period (1000 AD to 1800 AD), corresponding to the Islamic Age in South Asia.

A DATES model is considered valid if it meets the following criteria: Z-value > 2, nrmsd < 0.7, smoothness of the graph, and the mean date ± 2 standard errors is a valid date range. Our model passed these thresholds, yielding a mean date of admixture of 1363 AD, with a 95% confidence interval ranging from 1065 AD to 1660 AD.

Interestingly, these admixture dates align with historical accounts of the expansion of the Delhi Sultanate in the plains, as well as the Y-haplogroup TMRCA calculation for South Asian Syeds. Early Islamic migrations began around 1030 AD with the Ghaznavid Sultanates, and this wave of movement continued until the decline of the Mughal Empire in the late 17th century.

Figure 7: Date of West Asian admixture for UP-Bihar Syeds

Figure 8: (*Update) Analysis of 7 UP and 7 Bihar individuals shows UP Syeds have more recent West Asian ancestry (late 1400s) versus Bihar (1300s), with a combined average of 1467 AD. Notably, the earliest admixture signal (1259 AD) appears just after the Mongol conquest of Baghdad (1258AD).

Cross-checking using a PCA plot

We aimed to see how Syeds from UP-Bihar compared to their Indian geographical neighbours (Brahmins, Bhumihars, Kayasthas, Kshatriya, Kols) on a 2D PCA plot.

The results showed that the Syed cluster was pulled away from their geographical neighbours, shifting towards the West Asian group in the upper right corner, near populations like Levantines, Iraqis, and Saudis. In contrast, Brahmin groups, due to their high Steppe (Indo-European) ancestry, were pulled downward toward other high Steppe groups like Tajiks and Pashtuns. Interestingly, Knanaya Christians, a group known for their Assyrian origins (with about 10-15% West Asian admixture and no Steppe ancestry), were also pulled towards the top right, away from the other Indian groups.

Figure 9: PCA plot shows a pull of Syed cluster towards the top right away from their geographical neighbours

Figure 10: High Steppe Brahmin groups have a pull towards bottom to Tajiks and Pamiris, while Syed cluster is pulled towards top right. Same is the case for Knanaya Christians.

Generic Distribution of Haplogroups in the Non-Sampled Dataset

The reader might be curious about the general haplogroup distribution within this population. A randomly sampled distribution from one UP-Bihar Syed family looked like this, and could serve as a rough proxy for a broader, more generalized distribution:

Figure 11: A proxy haplogroup distribution in UP-Bihar Syeds

Some Commercial G25 Model Results

While I don’t endorse any particular company, I find it useful to reference some commercial results to demonstrate that they align with our qpAdm models. These G25-based models consistently detect around 5-10% Iraqi, Levantine, or West Iranian admixture, reinforcing the findings of our analysis. For any data science enthusiast, this convergence of results from multiple methods and approaches is a clear validation of our work—showing that different techniques point to the same underlying genetic story.

Future work

Based on our experience working with 21 samples, the results looked interesting in one way or the other. We are very interested in expanding our crowd-sourced experiment. If you belong to one of these South-Asian Syed families, we encourage you to take a test from one of the take home services, or reach out to our friends at South Asian Syed Y-DNA Study for any help.

References

1. Banu Hashim project

2. yFull for South-Asians under J-L859

3. Harney et al Assessing the performance of qpAdm: a statistical tool for studying population admixture

4. Narasimhan et al The Formation of Human Populations in South and Central Asia

5. Reich lab 1240k dataset

6. DATES tool code

7. Chintalapati et al The spatiotemporal patterns of major human admixture events during the European Holocene

8. Sample qpAdm output for right and left populations used, and setup example: