|
Post by Admin on Nov 6, 2021 20:31:01 GMT
Chromosome-wide ancestry proportions Beyond mean segment lengths, the proportion of ancestry per chromosome that descends from each ancestral population is also informative on the time of admixture [36, 37], since the longer the time after admixture, the smaller its variance [35]. While ancestry proportions contain less information than segment lengths, they are potentially more robust to misidentification of the segments boundaries. Building on models from refs. [35, 38, 39], we derived a new analytical expression for the distribution of ancestry proportions (for either phased or unphased data) given the initial admixture proportions and admixture time (Methods). This led to a maximum likelihood estimator of the admixture time and the initial proportions. For admixture between highly diverged populations, the method is expected to work well for intermediate admixture times (e.g., 10<t<100 generations [40]), as we demonstrated using simulations in which the true segment boundaries were known (S2 Fig). To apply our method to AJ, we used the LAI results and summed up the lengths of European and Middle-Eastern segments. However, our simulations showed that for Southern EU/ME admixture, the correlation between true and inferred ancestry proportions is only r2 ≈ 0.11 (S3 Fig), and therefore, we could not directly apply our method. To correct for the distortion of the distribution due to local ancestry inference, we again used EU/ME admixture simulations, and matched the variance of the AJ distribution to that of genomes simulated under admixture times between 10 to 60 generations. We found that the best fit to the AJ data, given a 4-way admixture model (Middle-Eastern, Southern EU, Eastern EU, and Western EU with proportions 50:34:8:8 (%), respectively) was obtained with admixture time of 32 generations (Fig 4) (95% bootstrapping confidence interval [31,37] generations), close to the time inferred above using the mean segment lengths. Fig 4 The Probability Density Function (PDF) of ancestry proportions in AJ and in simulations. The ancestry proportions in AJ were computed using LAI (RFMix). Simulations are of 1000 genomes with a history of an admixture pulse 32 generations ago between Middle-Eastern, Southern EU, Eastern EU, and Western EU populations. The density was estimated using a normal kernel. The admixture time was estimated by fitting the average standard deviation of the ancestry proportions across all chromosomes to the AJ data, where each chromosome was weighted by the square root of its length in cM. The confidence interval ([31,37] generations) was obtained by resampling AJ individuals, with replacement, 1000 times. The number of admixture events In light of identifying multiple EU ancestral sources, the assumption of pulse admixture might be unrealistic. In S1 Text section 6, we analytically derive the distribution of segment lengths and ancestry proportions for a double admixture model, where the initial admixture event was followed by a second contribution from one of the sources. However, we observed that the ancestry proportions from this model can sometimes be fitted well by pulse admixture. Given this and the considerable noise introduced by LAI, directly estimating the parameters of multiple admixture events is unlikely to be reliable. To overcome this problem, we first note that the inferred single admixture time still imposes some constraints on the admixture times and proportions in a double admixture model (Methods). Additionally, we notice that the estimated admixture time (≈30–35 generations) is very close to the time estimated for the AJ bottleneck event [9, 16]. If indeed two distinct admixture events have occurred, the single estimated admixture time represents a weighted average of the times of the two events (Methods). For that weighted average to coincide with the AJ bottleneck, it is reasonable to assume that one event has pre-dated the bottleneck, while the other has post-dated it, or at least that the two events have occurred at different stages of the bottleneck. This is expected to leave different traces when examining the ancestry of genomic segments with origin at around the time of the bottleneck, compared to the rest of the genome. We apply these insights in the following section.
|
|
|
Post by Admin on Nov 6, 2021 22:18:37 GMT
The ancestry of Identical-By-Descent (IBD) segments A number of recent studies have shown that sharing of identical-by-descent (IBD) segments is abundant in the AJ population, and is likely due to a severe bottleneck around 30 generations ago [4, 7, 9, 15, 16]. An open question is the relative timing of the bottleneck and the European gene flow, with our current and past [9] point estimates dating admixture at around or slightly earlier than the bottleneck. Given that most IBD segments but the very long ones (e.g., of length >7cM) coalesce around the time of the bottleneck, we contrast three hypotheses. If admixture completely predated the bottleneck, then IBD segments should have the same EU/ME ancestry proportions as observed genome-wide. If European admixture completely post-dated the bottleneck, then IBD segments should show exclusive ME ancestry. If, on the other hand, European gene flow occurred both before and after the bottleneck, then IBD segments should show an elevated (though not exclusive) ME ancestry compared to the rest of the genome [41–43]. Further, IBD segments of different lengths shared between AJ and other populations could shed light on the geographic origin of each admixture event. We detected long (>3cM) IBD segments using Germline [44] and Haploscore [45] (Methods). For segments shared within AJ individuals, we then computed the total amount of genetic material in IBD segments associated with each pair of diploid ancestries, namely, the fraction of SNPs in IBD segments where each of the two individuals sharing the segment has either homozygous EU ancestry, homozygous ME ancestry, or heterozygous ancestry. Clearly, errors in IBD segment detection and local ancestry inference could severely bias the conclusions of such an analysis. Fortunately, we could naturally account for these errors using the observed amount of genetic material in IBD segments shared between individuals labeled homozygous ME and homozygous EU, since the proportion of such segments is a direct measure of the noise level (Methods and S1 Text section 4). Our results demonstrate an over-representation of Middle-Eastern IBD segments, consistent with two waves of gene flow. Specifically, we estimated the European fraction of the AJ ancestry at the bottleneck as 42%, less than the 53% observed genome-wide (Methods). The contribution of post-bottleneck European gene flow required to explain these figures is 19% of the AJ ancestry (Methods). Considering only segments of length between [3,7]cM (as longer segments may descend from ancestors even more recent than the bottleneck) slightly increased the inferred magnitude of post-bottleneck gene flow to 22%, or 23% when considering only segments between [3,4]cM. Given a history of multiple admixture events, a natural question is the geographic source of each event. According to the documented AJ migration history, we speculated that the Southern-European gene flow was pre-bottleneck and that the Western/Eastern European contribution came later. Indeed, we note that the estimated proportion of ≈20% post-bottleneck replacement is close to our above estimate of ≈16% EU gene flow from sources other than Southern-EU as well as to TreeMix’s and Globetrotter’s results below (and perhaps also with our previous estimate of ≈15% EU ancestry based on AJ and Western European (CEU) data alone [46]). To test this hypothesis, we considered the European ancestry of IBD segments longer than 15cM, which are highly unlikely to predate the bottleneck. The proportion of AJ chromosomes with all regions masked but the >15cM IBD segments inferred by our geographic localization pipeline to be most likely Southern European decreased by 14.8% points compared to the genome-wide results. In contrast, the proportion of AJ chromosomes inferred to be most likely Eastern and Western European increased by 10.2 and 4.5% points, respectively. As a control, when we considered AJ individuals reduced to IBD segments of any length, there was no noticeable change. We also considered IBD segments shared between AJ and other populations (Fig 5), and observed that the number of segments shared between AJ and Eastern Europeans was ≈6-fold higher than shared between AJ and Southern Europeans (consistent with [5]), with this ratio increasing to ≈60-fold for segments of more recent origin (length >7cM). Further, the number of segments shared with Eastern Europeans was ≈2-fold higher than with Western Europeans or the people of Iberia (P = 5∙10−3 for the difference, using permutations of the EU regional labels), pointing to Eastern Europe as the predominant source of recent gene flow. Fig 5 The number of IBD segments shared between Ashkenazi Jews (AJ) and other groups of populations. IBD segments were detected by Germline and Haploscore, as explained in Methods. The population groups are as in Table 1. Note the different scale of panels (A) and (B) (segments of length between [3,7]cM and >7cM, respectively), and that sharing between AJ and either Southern Europeans or Middle Easterners completely vanishes for the longer (more recent) segments, indicating a relatively older divergence/gene flow. Also note that while sharing with Eastern Europeans is high compared to other groups, it is nevertheless a relatively rare event (≈0.04 segments per pair of individuals), in particular compared to sharing within AJ (≈3.4 segments per pair). Inferring the time and source of gene flow using additional methods Decay of admixture linkage disequilibrium (Alder), f4 statistics, and tree structure (Treemix) Refs. [47–49] have shown that linkage disequilibrium (LD) in an admixed chromosome, weighted properly, decays exponentially with genetic distance, and the Alder package was implemented to infer the admixture time and the ancestral sources. The admixture time inferred by Alder for AJ is broadly consistent with the LAI-based results, at 30–40 generations ago (Table 2; the P-value for admixture was significant under all tests). Table 2 Inferring the AJ admixture time and sources using Alder. Admixture times are in generations. The parameters were inferred, for each European region, using Alder’s self-determined minimal distance cutoff (rightmost column), above which the admixture LD decay is fitted. The other reference panel was always Middle-Eastern. Admixture time Amplitude Z-score Cutoff Southern Europe 39.8 2.8∙10−6 15.2 1.4cM Eastern Europe 29.6 8.6∙10−6 18.1 1.9cM Western Europe 35.3 8.2∙10−6 26.6 1.5cM For a simple admixture history, the LD curve amplitude increases as the reference population becomes closer to the true ancestral source. The Alder results (Table 2) would thus suggest that Eastern and Western Europeans are closer to the source of European gene flow into AJ, in contrast to the LAI-based results. However, when we ran Alder on simulated genomes with an admixture event, 30 generations ago, between Levant and Southern/Eastern/Western EU with respective ancestry proportions 50:35:12:3(%), the amplitudes were nearly identical to those of the real data, with the admixture times maintaining the same relative order and slightly overestimated at 34–41 generations ago. In fact, even simulations of pure Levant/Southern EU admixture resulted in higher Western/Eastern EU amplitudes than Southern EU. We thus conclude that, perhaps due to the complex admixture history in Southern Europe, Alder cannot infer the true ancestral sources, and that the results are still consistent with a model of predominantly Southern European contribution. A similar situation was observed when inferring the ancestral tree topology using f4 statistics [48] (S4 and S5 Figs) and TreeMix [50] (S6 Fig), both of which rely on the covariance of allele frequencies (or frequency differences) across populations. We measured the f4 statistic for the configuration (X,YRI;AJ,ME), where we used Yoruba (YRI) as an outgroup, and substituted different European regions for X (S4 Fig, part A). The European region that gave the highest value of f4, Eastern Europe (closely matched by Western Europe), is theoretically the one closest to the source of European gene flow. However, simulations with a dominant (and even exclusive) Southern European source resulted in highest f4 values for Eastern Europe as well. [This discrepancy might be explained, at least partly, by a strong Middle-East to Southern EU migration event [51] (S5 Fig), or by the small component of African ancestry in Southern Europeans [49].] Therefore, these results are still consistent with a dominant Southern EU source for AJ. We used the f4 statistics to infer the fraction of European ancestry in AJ, as explained in Patterson et al. [48]. Assuming that the true source is Southern Europe, the EU ancestry proportion is theoretically given by f4(West-EU,YRI;AJ,ME)/f4(West-EU,YRI;South-EU,ME)≈67% (S4 Fig, part B). However, when simulating genomes with 50% European ancestry, the f4-inferred fraction came out as 63%; thus, an inferred European ancestry proportion of 67% is broadly consistent with the RFMix-based estimate of ≈53%. We next ran TreeMix on AJ, Middle-East, the four European regions (West/East/South/Iberia), and YRI as an outgroup. The inferred tree (S6 Fig) suggests that AJ split first, followed by Middle-Easterners and Europeans. TreeMix then predicted replacement of ≈42% of the Southern EU ancestry by Middle-Eastern migration and ≈17% of the AJ ancestry by Eastern European migration, with the only other significant migration events coming from YRI and having much lower magnitude. However again, simulations with a predominantly Southern European ancestry yielded nearly identical results (S6 Fig). Interestingly, in simulations, TreeMix correctly estimated ≈13–14% Eastern EU ancestry in AJ when the true value was 12%, and almost no Eastern EU ancestry (≈2%) when none was simulated alongside Southern EU and ME ancestry; however, Eastern EU ancestry was erroneously estimated when the true simulated ancestry alongside Southern EU and ME was Western EU (16%). To summarize this section, we demonstrated that the raw results returned by Alder, f-statistics, and TreeMix must only be interpreted in light of simulations. Using simulations, the results were overall consistent with our model of an admixture event ≈30–35 generations ago in Southern Europe, with minor contributions of either Western or Eastern Europe.
|
|
|
Post by Admin on Nov 7, 2021 3:36:53 GMT
GLOBETROTTER analysis Finally, we considered GLOBETROTTER [21], which can infer both the contribution of each ancestral source and the admixture time. The first step in a GLOBETROTTER analysis is running CHROMOPAINTER [20], in order to determine the proportion of ancestry of each individual that is “copied” from each other individual in the dataset. Then, an ancestry profile for each population is reconstructed, representing the contribution of each other population to its ancestry [21, 22]. The inferred ancestry profile for AJ was 5% Western EU, 10% Eastern EU, 30% Levant, and 55% Southern EU. The combined Western and Eastern EU component is in line with our other estimates, as well as the dominance of the Southern EU component. However, the overall European ancestry, ≈70% (or ≈67% after calibration by simulations; S1 Text section 5), is about 15% higher than the LAI-based estimate, as well as our previous results based on whole-genome sequencing [9]. Our detailed simulations (S1 Text section 5) demonstrate that evidence exists to support either estimate. Possibly, the true fraction of EU ancestry is midway around ≈60%. Using the ancestry profiles calculated in the first step, GLOBETROTTER is also able to infer the admixture time and proportions, by assuming that the source groups could themselves be mixtures of the populations in the sample. A single admixture event was inferred for AJ (S1 Text section 5), where the first source, comprising 36% of the total ancestry, was 46% Western EU and 53% Eastern EU. The second source (64% of the total ancestry) was 35% Southern EU and 65% Levant, and the inferred admixture date was 34 generations ago. Our simulations (S1 Text section 5) show that the inferred genome-wide proportion of ≈22% Southern EU ancestry was significantly underestimated (by ≈20%-points) but that the overall inferred EU ancestry (here ≈58%) was accurate. The inferred admixture time was overestimated by ≈10 generations, implying an AJ admixture time 24 generations ago. With these adjustments, the results are broadly consistent with our conclusions so far. However, it remains open to explain the discrepancy between the inferred proportions from the ancestry profiles and the inferred proportions when running the full GLOBETROTTER pipeline. Bounding possible historical models We have so far provided multiple estimates for the ancestry proportions from each source and the time of admixture events. We now attempt to bring these estimates together into a single model and provide bounds on the model’s parameters. The results of all analyses (at least once examined in the light of simulations) point to Southern Europe as the European source with the largest contribution. At the same time, relatively large contributions from Western and/or Eastern Europe were also detected, with some analyses (IBD within AJ and between AJ and other sources, and GLOBETROTTER) showing stronger support for an Eastern European source. Based on historical plausibility, these admixture events must have happened at different times, implying multiple events. The inferred admixture time, when modeled as a single event, was between 24–37 generations ago across the methods we examined (corrected mean segment length and ancestry proportions, Alder, and GLOBETROTTER), very close to the time of the AJ bottleneck, previously estimated to ≈25–35 generations ago [9, 16]. Therefore, it is plausible to argue that one admixture event occurred before or early during the bottleneck, while the other event happened after the bottleneck, with the IBD analysis suggesting that the more recent admixture was with Eastern Europeans. Based on these arguments, we propose that a minimal model for the AJ admixture history should include substantial pre-bottleneck admixture with Southern Europeans, followed by post-bottleneck admixture on a smaller scale with Western or (more likely) Eastern Europeans. The estimates for the total European ancestry in AJ range from ≈49% using our previous whole-genome sequencing analysis [9], to ≈53% using the LAI analysis here, and ≈67% using the calibrated Globetrotter analysis. The proportion of Western/Eastern European ancestry was estimated between ≈15% (Globetrotter and the LAI-based localization method), and, if identified as the source of the post-bottleneck admixture, 23% (the IBD analysis). Therefore, the proportion of the Southern European (presumably pre-bottleneck) ancestry in AJ is between ≈26% to ≈52%, corresponding to [34,61]% ancestry at the time of the early admixture. Given these bounds, along with the admixture time estimate based on a single event (24–37 generations ago), we derived a constraint on the admixture times of the pre- and post-bottleneck events (Methods). We further assumed that post-bottleneck admixture happened at most 20 generations ago, when the effective population size has already recovered from the bottleneck (since our estimate of the post-bottleneck admixture proportions relied on the part of the genome not shared IBD; see the IBD analysis above and Methods). Finally, we assumed that post-bottleneck admixture happened no more recently than 10 generations ago, since no mass admixture events are known in the past 2–3 centuries of AJ history [52]. The results (Fig 6) show that given these constraints, the pre-bottleneck admixture time is between 24–49 generations ago. Fig 6 The relationship between the two admixture times in the Ashkenazi history, given bounds on the other admixture parameters. In the model, two populations (A and B) mixed t1 generations ago (early event; the proportion of ancestry contributed by population A, q, is indicated in the title of each panel). At a more recent time, t2 generations ago (recent event), migrants from A replaced another proportion μ of the admixed population (also in the titles). In each panel, we assumed that q and μ are known, as is the admixture time inferred under the assumption of a pulse admixture model (titles). Under these assumptions, and using Eq (5) in Methods, we plotted the time of the early event (t1) vs the time of the recent event (t2; blue circles). The weighted average (dashed lines) is a simple approximation, in which the time inferred under the pulse model is an average of t1 and t2, weighted by the admixture proportions q and μ, respectively. In the context of the Ashkenazi Jewish admixture history, population A is European and B is Middle-Eastern. Panels (A)-(D) represent the bounds on (i) the admixture time inferred under a pulse model (24–37 generations ago); (ii) the admixture proportions at the early and recent events (34–61% and 15–23%, respectively); and (iii) the time of the recent admixture event (10–20 generations ago). These bounds are justified in the main text. The results demonstrate that (i) the weighted average is a reasonable approximation, though the pulse admixture time is influenced more by the early event, perhaps as it results in more admixture tracts; and (ii) the most extreme values of the early AJ event are 24 and 49 generations ago. The lower bound corresponds to the lowest value of the inferred (single event) admixture time, the highest value of the time of the recent admixture event, and the largest contribution of the early event to the overall admixture proportions, and vice versa for the upper bound. Fig 7 A proposed model for the recent AJ history. The proposed intervals for the dates and admixture proportions are based on multiple methods, as described in the main text.
|
|
|
Post by Admin on Nov 7, 2021 22:10:26 GMT
Discussion Summary and lessons The ethnic origins of Ashkenazi Jews have fascinated researchers for over a century [53, 54]. The availability of dense genotypes for hundreds of AJ individuals, along with the development of new analysis tools, demonstrated genetic relatedness between AJ and other Jewish groups, and suggested Europe and the Middle-East as putative ancestral sources [4–8, 24]. Here we attempted, for the first time, to create a detailed portrait of the admixture events experienced by AJ during their dwelling in Europe. To this end, we used previously generated genome-wide array data for AJ, European, and Middle-Eastern populations (Table 1), as well as a variety of current and newly developed population genetics methods.
Before discussing the historical implications of our results, we point out two general lessons that emerge from the analysis. The first is that AJ genetics defies simple demographic theories. Hypotheses such as a wholly Khazar, Turkish, or Middle-Eastern origin have been disqualified [4–7, 17, 55], but even a model of a single Middle-Eastern and European admixture event cannot account for all of our observations. The actual admixture history might have been highly complex, including multiple geographic sources and admixture events. Moreover, due to the genetic similarity and complex history of the European populations involved (particularly in Southern Europe [51]), the multiple paths of AJ migration across Europe [10], and the strong genetic drift experienced by AJ in the late Middle Ages [9, 16], there seems to be a limit on the resolution to which the AJ admixture history can be reconstructed.
The second lesson is the importance of evaluating the results of off-the-shelf tools using simulations when studying closely related populations. When simulating relatively old (≈1k years ago) Middle-Eastern and European admixture (particularly Southern European), we found many tools to be of limited utility (see, e.g., the section on Alder, f-statistics, and TreeMix and S1 Text sections 1 and 2 on LAMP and PCAMask). Further, while we eventually were able to extract useful information off RFMix’s local ancestries, the raw results were not very accurate: the accuracy per SNP was only ≈70%, the mean segment length was more than twice than expected, and the variance of the ancestry proportion per chromosome was overestimated. When jointly analyzing LAI and IBD sharing, the inferred proportion of IBD segments that were either not IBD or had a random ancestry assignment was as high as ≈35% ((1-λ) in Methods), although fortunately, we were able to account for that in our model. We note, though, that problems of this nature are not expected for recent admixture events between more diverged populations.
Historical model and interpretation Our model of the AJ admixture history is presented in Fig 7. Under our model, admixture in Europe first happened in Southern Europe, and was followed by a founder event and a minor admixture event (likely) in Eastern Europe. Admixture in Southern Europe possibly occurred in Italy, given the continued presence of Jews there and the proposed Italian source of the early Rhineland Ashkenazi communities [3]. What is perhaps surprising is the timing of the Southern European admixture to ≈24–49 generations ago, since Jews are known to have resided in Italy already since antiquity. This result would imply no gene flow between Jews and local Italian populations almost until the turn of the millennium, either due to endogamy, or because the group that eventually gave rise to contemporary Ashkenazi Jews did not reside in Southern Europe until that time. More detailed and/or alternative interpretations are left for future studies.
Recent admixture in Northern Europe (Western or Eastern) is consistent with the presence of Ashkenazi Jews in the Rhineland since the 10th century and in Poland since the 13th century. Evidence from the IBD analysis suggests that Eastern European admixture is more likely; however, the results are not decisive. An open question in AJ history is the source of migration to Poland in late Medieval times; various speculations have been proposed, including Western and Central Europe [2, 10]. The uncertainty on whether gene flow from Western Europeans did or did not occur leaves this question open.
Caveats The historical model we proposed is based on careful weighting of various methods and simulations, and we attempted to account for known confounders. However, it is possible that some remain. One concern is the effect of the narrow AJ bottleneck (effective size ≈300 around 30 generations ago [9, 16]) on local ancestry inference and on methods such as TreeMix and f-statistics. We did not explicitly model the AJ bottleneck in our simulations, though a bottleneck may have been artificially introduced since the number of independent haplotypes from each region used to generate the admixed genomes was very small. However, as we discuss in Methods, this is not expected to affect local ancestry inference, since each admixed chromosome was considered independently. Another general concern is that while we considered multiple methods, significant weight was given to the LAI approach; however, this may be justified as the LAI-based summary statistics were more thoroughly matched to simulations. Another caveat is that our estimation of the two-wave admixture model is based on heuristic arguments (the multiple European sources and the differential ancestry at IBD segments), and similarly for the admixture dates. The IBD analysis itself relies on a number of assumptions, most importantly that the error in LAI and in IBD detection is independent of the ancestry and that most of the moderately long IBD segments descend from a common ancestor living close to the time of the bottleneck (see S1 Text section 4 and S7 Fig).
A general concern when studying past admixture events is that the true ancestral populations are not represented in the reference panels. Here, while our AJ sample is extensive, our reference panels, assembled from publicly available datasets, are necessarily incomplete. Specifically, sampling is relatively sparse in North-Western and Central Europe (and particularly, Germany is missing), and sample sizes in Eastern Europe are small (10–20 individuals per population). In addition, we did not consider samples from the Caucasus (however, this is not expected to significantly affect the results [5]). We also neglected any sub-Saharan African ancestry, even though Southern European and Middle-Eastern populations (including Jews) are known to harbor low levels (≈5–10%) of such ancestry [49, 56]. Generally, bias will be introduced if the original source population has become extinct, has experienced strong genetic drift, or has absorbed migration since the time of admixture. Additionally, a reference population currently representing one geographic region might have migrated there recently. We note, however, that as we do not attempt to identify the precise identity of the ancestral source, but rather its very broad geographic region, some of the above mentioned concerns are not expected to significantly affect our results. Additionally, as we show in S1 Text section 3, our pipeline is reasonably robust to the case when the true source is absent from the reference panel. We note, though, that there may be other aspects of the real data that we are unaware of and did not model in our simulation framework that may introduce additional biases. Finally, we stress that our results are based on the working hypothesis that Ashkenazi Jews are the result of admixture between primarily Middle-Eastern and European ancestors, based on previous literature [4–8] and supported by the strong localization signal of the ME source to the Levant. Strong deviations from this assumption may lead to inaccuracies in our historical model.
Future work The admixture history of Ashkenazi Jews thus remains a challenging and partly open question. To make further progress, the natural next step is to use sequencing data. Whole-genomes are now available for several European populations (e.g., [57]) as well as for Ashkenazi Jews [9] and some Middle-Eastern groups [58]. The accuracy of LAI is expected to increase for sequencing data, as also noted for other analysis tools (e.g., [59]). Additionally, whole-genomes will make it possible to run analyses based on the joint allele frequency spectrum of AJ and other populations. In parallel, denser sampling of relevant European and Middle-Eastern populations (mostly from Central and Eastern Europe) will be required in order to refine the geographic source(s) of gene flow.
Beyond data acquisition, we identify three major methodological avenues for future research into AJ admixture. First, any improvement in the accuracy of local ancestry inference will translate into improved power to resolve admixture events. Second, methods will have to be developed for the inference of continuous and multi-wave admixture histories (e.g., [35]) under LAI uncertainty. At the same time, inference limits will have to be established for events temporally or geographically near, as we began to develop here (S1 Text section 6; see also [40]). Finally, one may use the signal in the lengths of IBD segments shared between AJ and other populations and within AJ to construct an admixture model (e.g., as in [60]), provided that we can reliably detect shorter segments than is currently possible.
|
|
|
Post by Admin on Nov 30, 2022 23:51:23 GMT
Circa the 14th century, Ashkenazi women in Erfut, central Germany, carried a breast and ovarian cancer-indicative BRCA 1 mutation in their DNA. This mutation is unfortunately all too common in their modern descendants’ genomes, which is just one genetic sign that not a lot has changed in the ensuing 700-plus years. According to research being hailed as “the largest ancient Jewish DNA study so far,” published Wednesday in the prestigious Cell science magazine, by the 14th century Ashkenazi Jews had already received most of their main sources of genetic ancestry. When compared with the DNA markers of modern Ashkenazi Jews, there have been few changes to the genome in the centuries that have followed. This is just one of the findings afforded by analysis of ancient DNA extracted from teeth taken from a Jewish cemetery that was excavated in a salvage operation conducted according to the wishes of the local Jewish community alongside rabbinic advisers. The skeletal remains were later reburied in a 19th-century Jewish cemetery in Erfut. In 2013, German archaeologists excavated a portion of the ancient Jewish graveyard of Erfurt ahead of a municipal construction project, uncovering some 47 medieval graves. It was just the kind of potential treasure trove of centuries-old DNA that co-authors Hebrew University Prof. Shai Carmi and Harvard University David Reich were looking for, and they began their study of the remains five years later. “This work provides a template for how a co-analysis of modern and ancient DNA data can shed light on the past,” said Reich in a press release. “Studies like this hold great promise not only for understanding Jewish history, but also that of any population.” Through careful analysis of DNA extracted from teeth from 38 individuals followed by a comparison of hundreds of thousands of genetic place markers in modern Ashkenazi genomes, an international team of over 30 interdisciplinary researchers found that the Jews of Erfut “were noticeably more genetically diverse than modern Ashkenazi Jews,” according to co-author Carmi. “An even closer inspection revealed that the Erfurt population was divided into two groups: one with more European ancestry compared to modern Ashkenazi Jews, and one with more Middle Eastern ancestry,” said Carmi. Following some three years of testing and analysis, much of which was conducted in technologically advanced clean rooms at Harvard University, the results also indicated that the “founder event” or “bottleneck” that is evident in modern Ashkenazi Jewry’s DNA predated the establishment of the Erfut community, potentially by a millennium. According to Carmi, some of the genetic diseases associated with modern Ashkenazi Jews, including BRCA 1 mutations and Tay Sachs Disease, point to an extremely small initial population; as it grew, “pathogenic variants that were carried by the founders became widespread.” Among the methodologies utilized to gain information from the ancient teeth, the scientists sent 10 samples for radiocarbon dating, which found all 10 lived between about 1270 and 1400 CE. They also checked dental isotopes to see if the individuals had grown up drinking the same water and concluded that some were in fact immigrants. The results were published in Cell in an article, “Genome-wide data from medieval German Jews show that the Ashkenazi founder event pre-dated the 14th century.” Rare opportunity The opportunity to study the DNA of a medieval community such as Erfut was just what Carmi and co-author Reich were hoping for, Carmi told The Times of Israel on Wednesday. There is some historical documentation of the migration patterns and persecution of medieval Ashkenazi populations. However, said Carmi, “Given that no DNA sequences existed for historical Ashkenazi Jewry, we sought to generate ancient DNA data for this population. Our hope was to fill the gaps in our understanding of Ashkenazi Jewish early history.” The central German city was a thriving Jewish center in the Middle Ages and boasts one of the oldest still-standing synagogues in Europe. The Jewish community settled there in the 11th century; a massacre decimated the community in 1349 but Jews lived in the area until a final expulsion in 1454. At this time, a granary was constructed on top of the graveyard, sealing in the remains of thousands of Jews. “Jews in Europe were a religious minority that was socially segregated, and they experienced periodic persecution,” said Harvard’s Reich in a press release. “Our work gives us direct insight into the structure of this community.” Among the excavated 47 graves were two small nuclear families, including children buried near their father who apparently died from a violent blow to his skull. Other more distant family members were also discovered through genetic testing. Carmi said some eight of the 33 viable individual samples were related and allowed that it is possible that the limited available sample means the results do not fully reflect the entire community of Ashkenazi Jewry. “As with other ancient DNA studies, our historical inferences are based on a single site in time and space. This implies that our data may not be representative of the full genetic diversity of early Ashkenazi Jewry, as we have indeed inferred,” write the authors in the study. At the same time, the study indicates that “Medieval Ashkenazi Jews are best viewed not as a single homogeneous community (as it came to be at the present), but as an ‘archipelago’ of communities, differentially affected by founder events and mixture with local populations,” according to a FAQ sheet prepared by Carmi. A further conclusion is that late medieval Ashkenazi Jewry already carried certain disease-causing variants that became increasingly common among Jews as the years went by.
|
|