Project News
I recently undertook an examination of R1b1*. In case anyone wants
to peer over my shoulders, here's what I have so far.
R1b1* is a paragroup, generally defined as having the SNPs known as
M343 and P25 but lacking the downstream SNPs(M18,M73,M269, and
M335). R1b1* is much less common in Europe than its descendant
subclade R1b1c.
Through a variety of search methods, and with the help of friends on
this list, I found thirty-six haplotypes which are conclusively or
almost certainly R1b1*. These haplotypes, along with their
associated ysearch IDs, can be viewed in PDF or XLS form:
<a href='http://vizachero.com/R1b1/R1b1Table.pdf'>http://vizachero.com/R1b1/R1b1Table.pdf</a>
<a href='http://vizachero.com/R1b1/R1b1Table.xls'>http://vizachero.com/R1b1/R1b1Table.xls</a>
It is likely that there are additional R1b1* haplotypes in ysearch,
but except in a few limited cases it is necessary to have tested 25
or 37 STRs before one can conclusively draw a distinction between
R1b1* and R1b1c and this necessarily limited the pool. In some cases
personal knowledge of SNP tests provided some insights, and very near
or identical matches with established R1b1* members allowed the
inclusion of some 12-marker haplotypes in this analysis.
For the most part, R1b1* can be detected rather easily because it
almost always presents with DYS438=11 in conjunction with DYS464a=12
(or sometimes 13). In virtually all cases, nearby haplotypes (e.g.
R1a or R1b1c) could be excluded by examining their haplotypes at
alternate markers (e.g. DYS385) or their genetic proximity to known
Q, R1a, or R1b1c haplotypes.
Thankfully, an analysis of the R1b1* haplotypes revealed substantial
structure in the extant R1b1* population. This structure is robust,
and I was able to independently reproduce several clusters using a
variety of methods including median-joining networks, parsimony
trees, and distance trees. The following link is an example of a
distance-based tree, with the clades color-coded to match the
previous table of haplotypes.
<a href='http://vizachero.com/R1b1/R1b1Tree.pdf'>http://vizachero.com/R1b1/R1b1Tree.pdf</a>
All of the haplotypes in the tree above are in haplogroup R1b1* with
the exception of XKNX6, which is in haplogroup R2 and was included as
an out-group to root the tree. The tree shown is a UPGMA (distance-
based) tree, constructed using the Fitch method in Phylip and drawn
using FigTree. The inferred branching order of the clusters is
schematic but is also the most parsimonious explanation of the data.
Further, the structure of R1b1* was such that I could manually
construct a phylogeny using only three or four STRs with very low
mutation rates: DYS388, DYS426, DYS454, and YCA II. For simplicity,
I'll describe the clusters using this manual method.
Starting at the root, R1b1* can be divided into two large subgroups.
One group has DYS454=11 and the other group has DYS454=12.
The DYS454=11 group can be further divided into three clusters, one
each represented by DYS388=12, DYS388=13, and DYS388=14. It would
appear that the DYS388=14 cluster is a subclade of the DYS388=13
cluster.
The DYS454=12 group can be further divided into two smaller groups
using DYS426: one smaller group has DYS426=11 and the other has
DYS426=12. Additionally, the DYS426=12 appears to have further
substructure, with cluster having had a RecLOH event affecting YCA
II. One cluster has YCA II=21,21 and the other cluster having YCA
II=21,24.
It is also possible to distinguish five of these six clusters from
each other using only YCA II marker. The five distinct alleles are
18-23, 19-22, 21-21, 21-23, and 21-24 and the structure presented by
these alleles perfectly corresponds to the structure established
using DYS388, DYS426, and DYS454.
In summary, the six identified clusters are:
Cluster 1 (purple): DYS426=12, DYS388=12, DYS454=11, YCA II=19-22
Cluster 2 (blue): DYS426=12, DYS388=14, DYS454=11, YCA II=18-23
Cluster 3 (green): DYS426=12, DYS388=13, DYS454=11, YCA II=18-23
Cluster 4 (yellow): DYS426=11, DYS388=12, DYS454=12, YCA II=21-23
Cluster 5 (orange): DYS426=12, DYS388=12, DYS454=12, YCA II=21-21
Cluster 6 (red): DYS426=12, DYS388=12, DYS454=12, YCA II=21-24
It is worth noting that these four markers are among the slowest
mutating markers in FTDNA's 37 marker panel. Even the fastest
mutating of the four, YCA II, has an estimated mutation rate of
0.00123 or 1 mutation in 813 generations (24,000 years). The average
mutation rate for these four markers is 0.000425, or 1 mutation in
2,300 generations (70,000 years). It is for this reason that I am
not surprised that there is so little variance on these markers
WITHIN each of the six clusters. Moreover, the fact that there is so
much variance on these markers ACROSS the six clusters suggests to me
that the age of R1b1 is much, much older than any of these six
clusters or than its two major subclades (R1b1b and R1b1c since none
of these subclades show any variance in the four markers at all. My
analysis would, I think, be consistent with a age for R1b1 of 20,000
or more years. A proper dating using STR variance is problematic,
given the systematic nature of the search I used to find these
haplotypes in the first place, so I have not presented it.
It is worth noting that most cluster show strong geographic
localization, some more concentrated in Eastern Europe and others
more concentrated in Western Europe. Several of the haplotypes from
Eastern Europe are members of Sean Silver's Jewish R1b project. Some
of these Eastern European clusters also show signs of bottlenecking
during the historic period, which may be of interest to researchers
interested in Jewish genealogy.
A few word about search methods: in most cases I started with a
handful of haplotypes that I knew or suspected to be R1b1* based on
SNP tests or the DYS464/DYS438 motif mentioned earlier. In one case,
I asked a member of one of the clusters (surname Lumsden, cluster 5)
with the suspected to confirm it using a DeepSNP test and they
willingly did so. I then searched for neighbors in ysearch.
I iteratively examined each member of each cluster in ysearch, to
ensure that: 1) I captured all nearby examples within R1b1*; and 2)
that I excluded all nearby examples without R1b1*. In nearly each
case, when considering 37 markers, each cluster spanned a genetic
distance of significantly less than 10 and the nearest non-R1b1*
neighbor was a genetic distance of 19 or more away. Some of these
clusters were closer to R1a and R1b1c than others, but I am confident
that I have made few errors of inclusion or exclusion. To my
knowledge, each cluster includes at least one known SNP-tested
individual, and no cluster has a neighbor within a GD of 19 (at 37
markers) that has been SNP tested as anything BUT R1b1*. I made no
effort to exclude haplotypes that were very closely related or shared
similar surnames.
It would be possible, I think, to effect a similar search at SMGF.org
but probably not at yhrd.org since the later database includes few of
the slow-moving markers needed to discriminate R1b1* from R1b1c or R1a.
In summary, I came away from this exercise with a few conclusions:
1) The age of R1b1 is much older than an analysis of R1b1c would
suggest, a conclusion I reach based on the genetic distance between
these six clusters. I mention this as a caveat to those who would
extrapolate variance-based ages for R1b1c further upstream.
2) It should be possible to hypothesize where in R1b1* the major
subclades branched away. It looks to me like R1b1b is most closely
related to clusters 5 and 6 whereas R1b1c is most closely related to
cluster 2, though I admit that this notion could use a little more
development. Those interested could compare the modals for R1b1b
(4FNSC) or R1b1c (55GU9) to these R1b1* haplotype clusters.
3) The current Y-tree could need a massive revision if SNPs are found
to correspond to some or all of these clusters. Based on the trees I
constructed, I think it is not at all unlikely that two or more non-
redundant SNPs could be interjected between P25 and M269 and/or
between P25 and M73. While it is possible that some of these
clusters may ultimately be shown to be brother clades to R1b1b and
R1b1c, I think it is more likely that one of these clusters could end
up being a parent clade.
4) Tests of SNPs for placement on the Y-tree should include a wide
variety of R1b1* haplotypes, and at least one from each haplotype
cluster identified here. It is possible that SNPs currently believed
to be redundant based on testing one or two R1b1* haplotypes could be
found to be discriminatory based on testing haplotypes from a distant
cluster.