Tuesday, July 28, 2015

Comparison of Online Y-STR Predictors (Petrejcíková et al.) [Review]

An interesting study was published in 2014 based on Slovak Y-STR samples testing for 12 microsatellite markers. The main scope of this paper appears to be the investigation of the efficacy of three publicly available Y-STR haplogroup predictors (Athey, Cullen and YPredictor in alphabetical order) based on these 12 Y-STRs. Study contents shown below.

Y-SNP analysis versus Y-haplogroup predictor in the Slovak population.
Petrejcíková E, Carnogurská J, Hronská D, Bernasovská J, Boronová I, Gabriková D, Bôziková A, Maceková S. Anthropol Anz. 2014;71(3):275-85.
Human Y-chromosome haplogroups are important markers used mainly in population genetic studies. The haplogroups are defined by several SNPs according to the phylogeny and international nomenclature. The alternative method to estimate the Y-chromosome haplogroups is to predict Y-chromosome haplotypes from a set of Y-STR markers using software for Y-haplogroup prediction. The purpose of this study was to compare the accuracy of three types of Y-haplogroup prediction software and to determine the structure of Slovak population revealed by the Y-chromosome haplogroups. We used a sample of 166 Slovak males in which 12 Y-STR markers were genotyped in our previous study. These results were analyzed by three different software products that predict Y-haplogroups. To estimate the accuracy of these prediction software, Y-haplogroups were determined in the same sample by genotyping Y-chromosome SNPs. Haplogroups were correctly predicted in 98.80% (Whit Athey's Haplogroup Predictor), 97.59% (Jim Cullen's Haplogroup Predictor) and 98.19% (YPredictor by Vadim Urasin 1.5.0) of individuals. The occurrence of errors in Y-chromosome haplogroup prediction suggests that the validation using SNP analysis is appropriate when high accuracy is required. The results of SNP based haplotype determination indicate that 39.15% of the Slovak population belongs to R1a-M198 lineage, which is one of the main European lineages.
[Abstract] [Direct Link]

Are They Really Comparable?
Although all three predictors returned similar efficacy rates (~97-99%), it should be noted the authors' chief divisions of interest appear to be the conventional subclade designations currently used in both literature and the genetic genealogy community (e.g. R1a1a-M198). The authors correctly state Y-SNP testing is paramount in definitively gauging subclade classifications, especially for lines substantially downstream of a given haplogroup's phylogeny.

The rest of this entry determines whether these calculators display any other features which may give aspiring researchers reasons to choose one over another.

Subclade Coverage
A substantial difference is observed between the three. Athey's output is oriented around 21 categories spread across most of the major clades/subclades, although haplogroups not commonly found in West Eurasia (e.g. A-D) are unrepresented. Cullen improves on this significantly with 86 subclades, with Y-DNA I receiving the most attention (R1b to a lesser extent), with some improvements, such as well as the inclusion of "A&B". YPredictor has the highest count, hosting over 100 subclades, with the majority found in Y-DNA haplogroups E, G, J, N and R. With the exception of Y-DNA M and S, all are accounted for here.

STR count
Athey is capable of handling 111 Y-STR's (21 and 27-STR versions also available) with the format being listed in either numerical or Family Tree DNA (FTDNA) order. Cullen accepts a maximum of 67 STR's. YPredictor houses approximately 82 STR's. As such, all three are capable of handling a considerable number.

All three predictors permit the use of batched data and provide different means of categorising the data as seen fit by the user. Instructions are adequately provided for all three as well. As a research utility, however, YPredictor stands out through its' custom YFiler iterations (widely-used format in population genetics publications concerning Y-STRs) and debug feedback before predictions are made by the calculator.

Computational Time
This varies based on the user's CPU processing time, as well as whether they are manually entering STR values or inserting batched data. As such, this probably shouldn't be a pertinent factor in deciding which calculator to use.

Output Information
All three produce similar information (subclade prediction with probability expressed as a percentage).

Before summarising these findings, it is worth noting that Athey's predictor precedes Cullen's and YPredictor. As such, any perceived deficiencies in subclade breakdown or functionality are likely a result of age. Athey's predictor was widely used in the past, irrespective of the current application rate.

All three predictors are of use to genetic genealogists. This entry concludes the following "idealised" purposes for each:

  • Athey - For users keen to utilise upwards of 111 FTDNA Y-STR's as cross-validation against the other two
  • Cullen - Best for those seeking refined Y-DNA I or R1b subclade predictions
  • YPredictor - Most versatile and research-friendly, best worldwide coverage of Y-DNA subclades

As such, the three calculators certainly are comparable for making basic Y-STR predictions for West Eurasians, but obvious differences exist with respect to non-West Eurasian subclade coverage.

If compelled to make a single choice, I would recommend Cullen first to genetic genealogists of Northwest European paternal heritage (given the high frequencies of Y-DNA's I and R1b). YPredictor would be the best choice for those belonging to subclades more common outside Europe. This also explains why it has been extensively used in this blog to date. Athey's function has otherwise been usurped by the other two. 

Thursday, July 9, 2015

Presenting Bakhtiari Uniparental Marker Data [Original Work]

Bakhtiari people (Google Search)
The Bakhtiari people are one of Iran's ethnic minorities. Inhabiting the Iranian plateau's southwestern portion, the Bakhtiari traditionally maintained a hierarchical social structure with a genealogical basis (with organisations or positions including rish safids, kalantars, khans and ilkhani) [1]. Historically, the Bakhtiari have played a role in several pivotal events leading up to the formation of the modern Iranian state [2].

In recent years, the Bakhtiaris have received additional attention in the literature with respect to ancestry. This has been achieved predominantly via uniparental markers (Y-DNA and mtDNA) and coincides with work addressing the genetic origins of other ethnic minorities in Iran. For instance, in 2012, Grugni et al. expanded our understanding of Iranian Y-DNA across the country through sampling almost 1,000 unrelated men across 15 distinct ethnic groups (previous entry).

In spite of such developments, however, the Bakhtiari have not received much attention in either the genetic genealogy community or the literature. This entry attempts to explore the available data and arrive at a stable set of results for this group.

Khuzestan province, Iran (Wikipedia)

Search engines were limited to PubMed and Google Translate. Search terms included "Bakhtiari", "Y-DNA", "Y-Chromosome", "mtDNA", "mitochondrial", "STR", "SNP", "HVR" and "Iran". No limit was placed on publication date. All mtDNA and Y-DNA data was compiled. Where Y-STRs are presented, these were run through Vadim Urasin's YPredictor (v1.0.3 offline version). A 70% prediction strength threshold was implemented. If the resulting data is sparse, novel ways of consolidating the information will have to be devised and explained during the course of this entry.

Search Outcomes
Three studies were found to contain Bakhtiari uniparental data, with one partially covering Bakhtiari mtDNA (Derenko et al. 2013 [3]) and two for Y-DNA (Nasidze et al. 2008 [4], Roewer et al. 2009 [5]). The Bakhtiari populations featured mostly reside in Izeh, Khuzestan province, Iran [3-5] with a single sample coming from Lurestan province, Iran [4].

mtDNA Results
Derenko et al. featured only two Bakhtiari samples. One belonged to mtDNA H*, which was also observed in several Persian (Kerman province) and Qashqai samples, alongside a single Armenian. [3] The only other sample was mtDNA U2d2, also found in a single Persian (Kerman province). The authors noted that the combined frequency of mtDNA's U2c and U2d in Iran were highest among the Persians nationwide (approaching 10%) [3]. However, given the absence of additional samples, no reasonable conclusions can be drawn from these results.

Nasidze et al. provides both frequency and HVR1 derived variance data on the Bakhtiari and Ahwazi Arab populations [4]. The Bakhtiari appear to chiefly belong to mtDNA haplogroups N, U, H, T and J (below).

mtDNA Frequency Data from Khuzestan province, Iran {Nasidze et al. 2008)

Unfortunately, further information on subclade breakdown is not provided. However, as concluded by the authors and is evident through frequency data, the mtDNA profile of the Bakhtiari is almost identical to the Ahwazi Arab sample. Additionally, Nasidze et al. note "considerable sharing of HV[R]1 sequences" between these two groups [4]. In tandem with the inferences described above through Derenko et al., it appears that significant matrilineal marker overlap does exists across the Iranian plateau.

Y-DNA Results
Nasidze et al. first published data on 53 unrelated Bakhtiari men [4]. Due to substandard Y-SNP genotyping, the only conclusions that may broadly be discerned is the Bakhtiari chiefly belong to Y-DNA haplogroups J2-M172 (25%) and G-M201 (15%) (Data Sink). In this respect, these results cannot give observers a reliable indication of the Bakhtiari Y-DNA profile. Roewer et al.'s data indicates that some number of Bakhtiari do share the same core 17 STR haplotypes among one another (e.g. J2a4, T*)  but do not with any other samples across the country [5].

One "quick and dirty" way of addressing this problem is by using the YFiler (17 STR) Bakhtiari haplotypes (Data Sink) from Roewer et al. to "recharacterise" the Nasidze data. This is deemed the most suitable option for two reasons:
1) Nasidze et al. has an adequate sample size (n=53) but inadequate Y-SNP genotype selection
2) Roewer et al. has an inadequate sample size (n=18) and no confirmed Y-SNP testing, but the YPredictor data should provide reasonable subclade determination with a 70% probability threshold in place

"Recharacterisation" is achieved by expressing the Nasidze et al. data by the predicted subclade information provided by the Roewer et al. SNP predictions proportionally. For example, Nasidze et al. found "DE-YAP" at 8%, with the Roewer et al. predicted results showing 5.6% each for "DE*" and "E1b1b1". As both these subclades are contained within the DE-YAP node, the original value is recharacterised as DE 4% and E1b1b1 4%. The outcome is presented numerically (Data Sink) and demonstrated below (values rounded down to fit to 100%):

Y-DNA J2a4 constitutes the largest subclade (22.1%), with H (10.8%), R1a1a (8.9%) and T* (8.5%) following. The results imply considerable Y-SNP diversity within the Izeh Bakhtiari.

These results are somewhat at odds with that suggested by the Roewer et al. figures, particularly the frequency of Y-DNA J2-M172 (50% in Roewer et al. vs. 25% in Nasidze et al.). The most likely basis for this is sampling bias, given the former only tested for 18 individuals. It should be noted that Y-DNA J-12f2 has been documented to have a major (>60%) presence in Southwestern Iran (Quintana-Murci et al. 2001) with the majority of this likely being represented by downstream J2-M172 subclades (as per Grugni et al. 2012). It is therefore plausible for some Bakhtiari groups to yield exceptionally high frequencies of Y-DNA J2-M172 (likely J2a4 subclade) with future testing. The breakdown shown above is also broadly in line with past data from Southwestern Iran (Grugni et al. 2012).

It must be cautioned that literal interpretation of these results (both subclade breakdown and numbers) are not advised due to the inaccuracies brought by the "recharacterisation" and the lack of Y-SNP confirmation in Roewer et al.

It should also be emphasised that, as a tribal group, the Bakhtiari have most likely undergone genetic drift in their uniparental markers over time. As such, the finding of ~10% Y-DNA H is not completely surprising. Whether these values will be substantiated in future work is an open question.

The current evidence does suggest that the Bakhtiari closely resemble and share heritage with their immediate neighbours matrilineally, resting upon a backdrop of some common mtDNA diversity across the Iranian plateau. Inferences beyond this point will fall towards the realm of speculation.

The situation appears somewhat inverted on the Y-DNA side, where non-existent Y-STR haplotype sharing is observed with other groups in the Iranian plateau. The "recharacterised" data gives us an approximate idea of what the Bakhtiari Y-DNA profile should look like if Nasidze et al. used a better Y-SNP genotype panel.

Other ethnic minorities in Iran have received consistent attention in this respect, such as the neighbouring Qashqai and Lurs (Farjadian et al. 2011). The paucity in Bakhtiari uniparental marker data indicates this is very much an area that needs immediate attention. An initial first direction for researchers is to sample at least 50 unrelated individuals from Izeh using a more conventional Y-SNP genotype panel. Additional clarity will be gained by testing further areas, as well as reconciling the Bakhtiari tribal structure with these outcomes.

A very special thanks to the user "J Man" from Anthrogenica for bringing this interesting topic to my attention.

[Edit 10/07/2015]: I have also learned while researching this topic that Dr. Ivan Nasidze unfortunately passed away in 2012. His work served as an important early foundation towards understanding the genetic constitution of Caucasian and Iranian populations. May he rest in peace.

1. Bakhtiari. Last Accessed 25/06/2015: http://www.everyculture.com/Africa-Middle-East/Bakhtiari.html

2. Study of the Qajar government policy at the case of Household Bakhtiari. Last Accessed 6/07/2015: http://waliaj.com/wp-content/2014/Issue%201,%202014/26%202014-30-1-pp.124-127.pdf 

3. Derenko M, Malyarchuk B, Bahmanimehr A, Denisova G, Perkova M, Farjadian S. Complete mitochondrial DNA diversity in Iranians. PLoS One. 2013 Nov 14;8(11):e80673. doi: 10.1371/journal.pone.0080673. eCollection 2013.

4. Nasidze I, Quinque D, Rahmani M, Alemohamad SA, Stoneking M. Close genetic relationship between Semitic-speaking and Indo-European-speaking groups in Iran. Ann Hum Genet. 2008 Mar;72(Pt 2):241-52. doi: 10.1111/j.1469-1809.2007.00413.x. Epub 2008 Jan 20.

5. Roewer L, Willuweit S, Stoneking M, Nasidze I. A Y-STR database of Iranian and Azerbaijanian minority populations. Forensic Sci Int Genet. 2009 Dec;4(1):e53-5. doi: 10.1016/j.fsigen.2009.05.002. Epub 2009 Jun 5.