Phylofriend User Guide

Dirk Struve
phylofriend at projectory.de
https://github.com/yogischogi/phylofriend/
March 20, 2018

1
Contents
1 Introduction 3

2 Installation 4

3 Command Line Options 5

4 Examples 6
4.1 Create a Phylogenetic Tree . . . . . . . . . . . . . . . . . . . . 6
4.2 Pimp Your Tree with Nicer Labels . . . . . . . . . . . . . . . . 6
4.3 Use a Specific Set of Mutation Rates . . . . . . . . . . . . . . 7
4.4 Calibrate Your Data . . . . . . . . . . . . . . . . . . . . . . . 7
4.5 Count Mutations . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.6 Marker Statistics . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.7 Use Data from YFull . . . . . . . . . . . . . . . . . . . . . . . 9
4.8 Extract Data from a Spreadsheet in CSV format . . . . . . . . 10

5 Technicalities 11
5.1 Source Code Documentation . . . . . . . . . . . . . . . . . . . 11
5.2 Mutation Model . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.3 Mutation Rates . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.4 CSV Input Format . . . . . . . . . . . . . . . . . . . . . . . . 13
5.5 Text Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.6 YFull Format . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.7 PHYLIP Format . . . . . . . . . . . . . . . . . . . . . . . . . 14

6 Theory 15
6.1 Haplogroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
6.2 Haplotypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
6.3 Phylogenetic Trees . . . . . . . . . . . . . . . . . . . . . . . . 16
6.4 Genetic Distance . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.5 Modal Haplotype . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.6 Age Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.6.1 Mutation Counting . . . . . . . . . . . . . . . . . . . . 20
6.7 From Distances to Trees . . . . . . . . . . . . . . . . . . . . . 22

References 26

2
1 Introduction
Phylofriend’s main purpose is to calculate genetic distances from Y-DNA
STR values. The results can be used as input for the PHYLIP[6] program
to create phylogenetic trees.
When I started creating phylogenetic trees I often found myself in a dif-
ficult position. As a Linux user I was missing some of the tools available
under Windows. So I started to write this program to fill in the gaps and
make myself comfortable again.
This does not mean that you can not use Phylofriend when working under
Windows or the Mac. But currently there is no binary distribution available
and you will probably face a hard time installing Phylofriend and the asso-
ciated programs. So I only recommend this if you are an experienced user.
Phylofriend has some nice features. It can be used

• to create phylogenetic trees using the PHYLIP[6] program. Y-STR

values from Family Tree DNA projects can easily be imported.

• as a programming library. Phylofriend is written in Google’s Go pro-

gramming language. This language is not only suited to solve Google’s
large scale programming problems. It is also an excellent tool for part
time programmers who have to concentrate on their projects (often
students).

• to extract STR values from Family Tree DNA projects and convert it
into simpler text files that are better suited for further processing.

• to automate phylogenetic tree creation. Phylofriend is a command line

tool and this scares many people away. But if you have to repeat the
same tasks over and over again you will eventually start to write some
scripts and this is where command line tools come in handy.

I hope this program will be useful. Have a good time!

Dirk

3
2 Installation
This guide is mainly targeted towards persons who use Linux Mint or other
Linux versions of the Debian family. Some familiarity with the use of Linux
commands is assumed.
Currently there are no binary distributions available for Windows or the
Mac. Users of these operating systems can use Phylofriend as well, but they
will experience some laborious installation work. The best way is to follow
the instructions provided on the Go home page and the PHYLIP home page.
The following list applies to Linux users only:

1. Make sure that the Go programming language is installed. If not it can

be installed by typing
sudo apt-get install golang

2. Read the Go Getting Started guide. Make sure to set your GOPATH
variable and include it in your PATH so that Go programs can be
found.

3. For the creation of phylogenetic trees install the PHYLIP program

package by typing
sudo apt-get install phylip

4. Fetch the Phylofriend program with

go get github.com/yogischogi/phylofriend

5. Install the program with

go install github.com/yogischogi/phylofriend

4
3 Command Line Options
Command line options may be given in arbitrary order.
-help Prints available program options.
-personsin Filename or directory of files containing the persons’ Y-STR
values. If this is a single file it must contain results for multiple persons.
The input file format is CSV (comma separated values) or text format.
If a directory is provided for input it must contain multiple files in
YFull format, each file containing the results for a single person. The
person’s ID is extracted from the filename.
personsin supports multiple file names separated by commas.
-labelcol Number of the column that is used for labels when reading CSV
files.
-mrin Filename of the mutation rates to use.
-model Mutation model to use. This may be hybrid or infinite. hybrid
uses uses stepwise counting for most markers except for the palindromic
ones. infinite uses the infinite alleles mutation model for all markers.
-anonymize If this is true persons’ names are replaced by numbers.
-modal Creates modal haplotype and performs TMRCA calculation.
-phylipout Filename for the distance matrix that can be fed into the PHYLIP[6]
program.
-mrout Filename for the output of the currently used mutation rates.
-txtout Filename for text output of persons and Y-STR values.
-htmlout Filename for HTML output of persons and Y-STR values.
-nmarkers Uses only the given number of markers for calculations.
-gentime Generation time.
-cal Calibration factor.
-reduce Reduces the number of persons by the given factor (for large num-
bers of samples).
-statistics Prints marker statistics.

5
4 Examples
4.1 Create a Phylogenetic Tree
1. Copy persons’ data from a Family Tree DNA project website into a
spreadsheet. If the Y-STR values do not appear properly try inserting
them into the spreadsheet as unformatted text.

2. Save the spreadsheet in CSV (comma separated values) format, for

example persons.csv.

3. Start a terminal or command line interpreter and go to the directory

where you stored persons.csv.

4. Create a matrix of genetic distances by typing

phylofriend -personsin persons.csv -phylipout infile
-mrin 67-average.txt
Here we use a set of average mutation rates for 67 markers. The file 67-
average.txt can be found in the directory phylofriend/mutationrates/.
The result is a matrix that contains the number of generations as a
measure for genetic distances between the persons.

5. Use the PHYLIP program to create a tree in Newick format with

/usr/lib/phylip/bin/kitsch
You will need to answer some questions. Usually the default values
are good enough. The results will be two text files, one named outtree
which contains the tree in Newick format and another one named outfile
which contains a more human readable description.

6. Create an image of the tree by typing

/usr/lib/phylip/bin/drawgram
Use outtree as the input file name. The resulting tree will be stored in
a file named plotfile.
A nice alternative to visualize the tree is the use of the Trex[3] web-
server. You can copy the contents of the file outtree into the Trex
window.

4.2 Pimp Your Tree with Nicer Labels

By default Phylofriend assumes that your persons input file’s first column
contains a list of IDs. This is usually a Family Tree DNA Kit number. The
resulting tree is hard to read. Many projects keep names in another column.

6
You can access this column by using the labelcol option. Suppose your second
column contains names. You can create a distance matrix with names instead
of IDs by typing
phylofriend -personsin persons.csv -labelcol 2
-phylipout infile -mrin 67-average.txt
Due to compatibility issues with other programs the labels must be 10
characters long and may only contain 8-bit characters. Phylofriend will apply
a transformation to make sure that the requirements are fulfilled but the
result is sometimes a bit strange.
You can also use the labelcol option to create trees that contain the origins
of people or the haplogroups. Although I strongly recommend to build trees
only from people who belong to the same haplogroup this is sometimes useful
if you want to know if different haplogroups are close on their Y-STR values.
If you want to publish your tree you will often need to protect the privacy
of the members. This is what the anonymize option is for. By typing
phylofriend -personsin persons.csv -phylipout infile -anonymize
-mrin 67-average.txt
you will get a distance matrix where the names are replaced by numbers.

4.3 Use a Specific Set of Mutation Rates

Phylofriend supports the use of arbitrary mutation rates by the mrfile option.
The phylofriend/mutationrates directory contains some files with mutation
rates. The average mutation rates where taken from [9]. If you like to
compare on 67 markers or 111 markers you can use
phylofriend -personsin persons.csv -phylipout infile
-mrin 67-average.txt
or
phylofriend -personsin persons.csv -phylipout infile
-mrin 111-average.txt

4.4 Calibrate Your Data

Mutation rates depend on the method applied to calculate genetic distances
and the sample populations used. Mutations themselves occur by coinci-
dence. Average mutation rates often yield acceptable results but in most
cases you will have to calibrate your data, especially if you want to calculate
genetic distances in years.
Phylofriend provides two options for data calibration: gentime, the gen-
eration time in years and cal an additional calibration factor. Internally

7
they are just multiplied together but using two separate factors seems more
convenient for typical use cases.
A generation time of 32 years has proven to show good results [1]. You
can use it by typing
phylofriend -personsin persons.csv -phylipout infile
-gentime 32 -mrin 67-average.txt
In reality an optimal result is often hard to achieve. Especially within the
range of genealogical time frames (about 400 years) and only small numbers
of persons who have tested, you are often left with a large statistical error.
Even worse, the method of Y-STR counting has numerous pitfalls, some of
which are discussed in [7].
If you have a reliable paper trail or a well defined historic event you
can calibrate your data using the cal option. With cal you just provide
an additional calibration factor that is multiplied to the calculated genetic
distances, for example
phylofriend -personsin persons.csv -phylipout infile
-gentime 32 -mrin 67-average.txt -cal 1.2
multiplies all genetic distances by a factor of 1.2.

4.5 Count Mutations

If you want to count mutational differences between persons, for example on
a 37 marker scale, you can achieve this by typing
phylofriend -personsin persons.csv -phylipout distancecount.txt
-nmarkers 37 -cal 37
The nmarkers option restricts all calculations to the given set of markers
(usually 37, 67 or 111). If a person has not tested for enough markers,
he is excluded from the calculation. Because Phylofriend uses average values
internally, all results are devided by 37 in the previous example. To get whole
numbers we must compensate for this effect. This is done by multiplying the
results with 37 using the cal option.
The internal use of average values seems complicated at first but it allows
Phylofriend to compare persons who have tested for different numbers of
markers.
Phylofriend’s results may also differ from other programs because Phylo-
friend uses it’s own hybrid mutation model.
An alternative approach to mutation counting is using a set of mutation
rates that was especially created for that purpose, for example:
phylofriend -personsin persons.csv -phylipout distancecount.txt
-mrin 37-count.txt

8
 The difference to the previous example is that the nmarkers option in the
previous example makes sure that all persons have tested for all 37 mark-
ers, while the use of the 37-count mutation rate also accepts persons who
have tested for a lower number of markers. The result is an estimate. This
option was introduced because next generation sequencing has revealed Y-
STR results for 587 markers. However due to the technical restrictions of
this method, different persons usually get results for different, but largely
overlapping, marker sets.

4.6 Marker Statistics

Phylofriend can give you detailed statistics about marker values. It calculates
the minimal and maximal value for each marker and the number of occur-
rences. You can activate the statistical output by using the -statistics=true
option, for example
phylofriend -personsin=s11481.csv -statistics=true -nmarkers=6
prints out the results for the first six markers (-nmarkers=6). In our example
it is assumed that the persons’ data is in a file called s11481.csv. This is the
output:

DYS393, Min: 13, Max: 13, 13:15,

DYS390, Min: 24, Max: 24, 24:15,
DYS19, Min: 14, Max: 14, 14:15,
DYS391, Min: 10, Max: 11, 11:14, 10:1,
DYS385a, Min: 11, Max: 12, 11:10, 12:5,
DYS385b, Min: 12, Max: 14, 12:1, 14:14,

Each line contains the marker name and the minimal and maximal values.
After that each marker value is listed together with it’s frequency, separated
by a colon.
Let’s take DYS391 as an example. It has two values, 10 and 11. The
value 10 occurs 1 time and the value 11 occurs 14 times.

4.7 Use Data from YFull

YFull reports up to 587 STR values (March 2018), but not all of them for
each person. A Family Tree DNA Big Y test usually reveals about 500 STR
markers, some times less, some times more.
To use these results put the YFull result files for every single person into
a directory, for example inputdir. Call Phylofriend by typing:

9
phylofriend -personsin inputdir -phylipout infile -modal
-mrin 587-count.txt
This command will use up to 587 markers for comparison. The effect of
the file 587-count.txt is that the mutation differences are just counted. An
average mutation rate for 587 markers is not available.

4.8 Extract Data from a Spreadsheet in CSV format

Spreadsheet data is often uncomfortable to handle, especially if you want to
write your own program and need to parse it. For this purpose Phylofriend
supplies the txtout option. It writes data to a text file in simplified form.
The easiest way to use it is
phylofriend -personsin persons.csv -txtout persons.txt
This extracts the data from persons.csv and writes it to persons.txt. The first
column of persons.txt contains the first column found in persons.csv, usually
a set of IDs. The following columns contain the Y-STR values. All columns
are separated by tabs.
If you want to use another column you can use the labelcol option:
phylofriend -personsin persons.csv -txtout persons.txt
-labelcol 2
This extracts the second column from persons.csv and writes it to the first
column of persons.txt.
When using the txtout option you will need to specify how many Y-STR
values are written to the text file. This is done by nmarkers. Phylofriend
will write the same number of Y-STR values to each line. Missing values are
written as 0, larger sets of Y-STR values are truncated. This is how to write
a full set of 111 markers:
phylofriend -personsin persons.csv -txtout persons.txt
-nmarkers 111

10
5 Technicalities
5.1 Source Code Documentation
To access the source code documentation point your web browser to:

• http://godoc.org/github.com/yogischogi/phylofriend

If you want to modify the source code it is best to use godoc locally on
your computer.

1. If godoc is not yet installed install it by typing

sudo apt-get install golang-go.tools
(Linux Mint)

2. Start the documentation web server with

godoc -http=:6060

3. Point your web browser to localhost:6060.

4. Click on Packages and search for phylofriend.

This will give you a nice overview of the internal program documentation.
You can also click on function names to browse the source code.

5.2 Mutation Model

The two basic mutation models are the infinite alleles model and the stepwise
mutation model as explained by Bruce Walsh[13]. Phylofriend uses a hybrid
mutation model. Most markers are calculated using the stepwise model,
palindromic markers are calculated as described in [4].
As the method for calculating the genetic distance is likely to change
with time please look at the internal program documentation if you need
more details.

5.3 Mutation Rates

The directory phylofriend/mutationrates contains files with predefined mu-
tation rates.

11
Average Mutation Rates
• 12-average.txt

• 37-average.txt

• 67-average.txt

• 111-average.txt

In these files average mutation rates are used for the corresponding set
of markers. The mutation rates for 12, 37, 67 and 111 markers were taken
from [9].

12 Marker Single Mutation Rates

• 12.txt

• 37-12.txt

• 67-12.txt

• 111-12.txt

These files are basically the same as the average mutation rates files but
for the first 12 markers the mutation rates are set per marker. This puts ap-
propriate weight on very stable markers. The mutation rates were taken from
Wikipedia[12]. Although there may be some doubt if data from Wikipedia
can be trusted the first 12 markers are well known and they are in use for a
long time. So I just adopted them. The values should be good enough for
most purposes.
For each file the 12 marker mutation rates were multiplied by a calibration
factor so that their average value matches that from the rest of the file.
These files are useful for deeper history or if you observe changes on long
time stable markers. If several persons share the same value on a very stable
marker they probably belong together. So try these mutation rates to see if
the results make more sense than those obtained by using average markers.

Mutation Rates for Marker Counting

• 12-count.txt

• 37-count.txt

12
 • 67-count.txt

• 111-count.txt

• 587-count.txt

These mutation rates can be used for marker counting between differ-
ent persons, even if they have not been tested on the exact same markers.
Phylofriend will give you an estimate about the expected marker difference
on the specified scale.
This approach is recommended for large marker sets from next generation
sequencing.

5.4 CSV Input Format

Example:

id1,"Dirk Struve",Germany,R1b-CTS4528,13,24,14,11,11-14,12,...
id2,"Pyl. O. Friend",Germany,R1b-CTS4528,13,24,14,11,11-14,...

When importing a file in comma separated values format the first column
must contain IDs. An arbitrary number of columns containing custom infor-
mation may follow. The last columns must contain at least 12 Y-STR values
in Family Tree DNA order. Rows containing comments are allowed.
Phylofriend will always try to parse the file as best as it can.

5.5 Text Format

Example:

Dirk_Struv 13 24 14 11 11 14 12 12 12 12 14 28
Pyl._O._Fr 13 24 14 11 11 14 12 12 12 12 14 28

The text format is a simplified format intended for easy parsing and to
work well with other programs. For compatibility reasons the first column is
exactly 10 characters long and contains only non Unicode characters. Spaces
are transformed into underscores. The following columns contain Y-STR
values separated by tabs.

13
5.6 YFull Format
YFull files have a name like STR for YF01234 20160222.csv. Each line of
the file contains the result for a single marker and sometimes additional
information. Example:

DYS390;24;
DYS391;11;
DYS392;14;?
DYS393;13;

Although the file is in CSV format, semicolons are used instead of commas
as separators. This may cause trouble if you try to add additional results
by hand. Phylofriend does not care if a marker is provided multiple times.
The last occurrence of a name is considered the valid one. Thus you can add
results from other testing companies just by adding them to the end of the
file.
Phylofriend tries to extract a person’s ID from the filename. If a file is
named STR for YF01234 20160222.csv, the ID will be YF01234.
If you want to provide your own files, you do not need to stick to the YFull
naming convention. Just use the desired ID as a filename like ID1234.csv,
but the file must end in .csv.

5.7 PHYLIP Format

Example:

2
Dirk_Struv 0 0
Pyl._O._Fr 0 0

The first line contains the number of entries. An entry line contains
an ID that is 10 characters long and contains only non Unicode characters.
Spaces are transformed into underscores. The columns containing genetic
distances are separated by tabs. For readability reasons Phylofriend writes
only integers. If you need more precision you can scale the distance by using
the cal option.

14
6 Theory
6.1 Haplogroups
Parts of the human Y chromosome are only inherited from father to son
[1, 5]. Usually these parts are unchanged but sometimes a mutation at a
single position occurs. Such mutations are called SNPs (Single Nucleotide
Polymorphisms). They are very stable and can be used to group people into
ancestral lines.

Figure 1: A set of single mutations defines a person’s haplogroup. Persons

who belong to the same haplogroup usually share a common ancestor within
a few thousand years.

Figure 1 illustrates a typical situation. Many Europeans belong to hap-

logroup R which is characterized by the mutation M207. The M207 marker is
inherited throughout all following ancestral lines. Later the mutations M420
and M343 occurred. They are mutually exclusive to each other thus defining
separate lineages. The marker M420 defines the haplogroup R1a which is
common in Eastern Europe and the marker M343 defines the haplogroup
R1b which is common in Western Europe. Because all people who belong to
R1a and R1b share the M207 marker we know that they also share a common
ancestor a long time ago.
SNPs occur approximately once in a hundred years [1] but usually only
SNPs that are older than 1000 years are listed at ISOGG [10]. Under normal
circumstances Persons who belong to the same haplogroup share a common
ancestor within a few thousand years. If you know your haplogroup it will
tell you something about your deep history (thousands of years). Newer tests
like Family Tree DNA’s Big Y [5] also make it possible to find relatives within
a genealogical time frame.

15
6.2 Haplotypes
Most people want to know more about family relationships. Relationships
are often verified by STR (Short Tandem Repeats) mutations. STRs are
repetitions of certain genetic patterns. They group people into haplotypes.

Figure 2: A person’s haplotype is defined by patterns of mutations. Persons

with the same haplotype usually share a common ancestor within a few
hundred years if they take a standard test with 37 markers or more.

On the Y chromosome there are many different patterns that repeat

themselves. Every pattern has a name, for example DYS393. If we count
the number of repetitions for a specific pattern we get a value, for example
DYS393=13. Figure 2 illustrates the situation for an exemplary marker. The
pattern is repeated 3 times. So this marker has a value of 3.
If a mutation occurs the number of repetitions changes. A single marker
mutates rarely but the modern standard tests use many markers at the same
time, most often 37, 67 or 111. The combination of all these markers defines
a haplotype. Persons who share the same haplotype usually also share a
common ancestor within a few hundred years.
It is important to realize that haplotypes are not as good as haplogroups.
Haplotypes sometimes overlap between separate lineages. So when working
with haplotypes the haplogroup should also be known.
Phylofriend uses haplotypes to calculate genetic distances between per-
sons by counting the number of different mutations.

6.3 Phylogenetic Trees

Phylogenetic trees group people together according to their genetic distances.
They give us a picture of how different persons are related.
Figure 3 shows some exemplary trees. If a father has two sons we would
expect that both sons are at a little genetic distance away from their father.
If they are not twins they also should be separated by each other. Tree a
illustrates this situation. This is exactly the tree we would get if we could
measure the genetic distance at a high resolution and know the birth dates
of father and sons. Because we know the birth dates we would put the father

16
Figure 3: Phylogenetic trees group persons together according to their
genetic distances. Genetic distances are often associated with time scales
but this is only true for long time spans.

before the sons and associate the horizontal axis with a time scale assuming
that genetic distance is proportional to time.
Reality however shows a different picture. The currently available stan-
dard tests (37, 67 and 111 markers) only offer a low resolution. So most often
we can not measure any genetic distance between father and sons. This is
illustrated by tree b. Because the tree illustrates genetic distances and not
time distances father and sons are all side by side.
If son 2 develops a mutation by chance we get a confusing picture. Son
2 is suddenly far away from his father and his brother. This is shown by
tree c. The reason for the great distance between father and son 2 is the low
resolution of the test. If we take the mutation rates from [9] we can calculate
how long it takes on average for a mutation to occur. For a 37 marker test
the result is 280 years, for a 67 marker test 210 years and for a 111 marker
test 130 years.
This means if the father and his sons took a 37 marker test and son 2 has
developed a mutation by chance he will appear to be 280 years away from
his father but the only reason for that is the low resolution of the test.
In most cases the standard tests are good enough. We do not need a
paternity test for genealogical purposes but it is important to remember that
the standard tests only places persons into time frames of hundreds of years.

17
6.4 Genetic Distance
There are many ways to calculate genetic distances. Here we use the method
of mutation counting as described by Anatole Klyosov in [8]. We just count
the number of mutation one persons differs from another and take the result
as the genetic distance. Take a look at these two 6-marker haplotypes:

DYS393 DYS390 DYS19 DYS391 DYS385a DYS385b

Carl 13 24 14 11 11 14
Clas 13 23 14 12 11 14

At DYS390 and DYS391 Clas differs from Carl by one mutation. Thus
their genetic distance on the 6-marker scale is 2.
To represent this genetic distance in years we need to know how often
mutations occur. Thus we need a mutation rate. The mutation rate is
defined as follows:
m
k= (1)
gy
k: Mutation rate per marker and generation
m: Number of mutations
g: Number of generations
y: Number of Y-STR values (markers)

Mutation rates are derived from sample populations. But mutations occur
by coincidence. Thus for a precise measurement we need large numbers of
mutations. Different family lineages often expose different mutation rates.
So the standard mutation rates should be considered as a first estimate. You
can use Phylofriend’s cal option to adjust your data to historic events.
Now we can use equation 1 to calculate a genetic distance in generations.
Formula 1 is equivalent to:
m
g= (2)
ky
This gives us the number of generations between Carl and Clas. The only
thing we still need to know is the generation time. For historical purposes a
generation time of 25 years is commonly used. We get the time by multiplying
the number of generations with the generation time.

t = gd (3)

18
 t: Genetic distance in years
g: Number of generations
d: Generation time in years

By substituting g with formula 2 we get the genetic distance in years:

md
t= (4)
ky
t: Genetic distance in years
m: Number of mutations
d: Generation time in years
k: Mutation rate per marker and generation
y: Number of Y-STR values (markers)

Let us try this out. The number of mutations between Clas and Carl is
2. The mutation rate for a 6-marker haplotype is 0.002 [8]. So
2 · 25
t= = 4167 years (5)
0.002 · 6
Wow, that is a very long time. But what does this number actually mean?
First, it is an average value and there is a very big margin of error to it but
it is useful as a first estimate and to group people together according to their
genetic distance. Second it is the time Clas and Carl would be separated if
Clas would be an ancestor of Carl (or the other way round). It is not the
time to their most recent common ancestor (TMRCA).
If Clas and Carl would be living today, how long would be the the time
to their most recent common ancestor? The first guess is that the common
ancestor would be in-between his descendants in terms of genetic distance.
So Claus and Carl would be 4167years/2 = 2085years away from him.
This is a good first guess but unfortunately reality is much more compli-
cated. Mutations occur by coincidence and due to the laws of statistics some
people develop only a few mutations while others develop much of them. So
our first guess may give a totally wrong impression. Generally it is not a good
idea to calculate the time to a common ancestor just based on the results
of two people. The best way is to identify a group of people who share a
common lineage and calculate the time to the most recent common ancestor
for the whole group. Anatole Klyosov describes how this is done in [8].
For those who still want to use a TMRCA value based on the results of just
two people Bruce Walsh has developed a method to give a time estimate[14].
This is better than the naive calculation we have used before but it is still
an estimate and it is only valid for demographically stable populations.

19
 The whole story is of course more complicated than the simple calculation
presented here. Different mutation models exist and some genetic markers
should be treated in a special way. Phylofriend uses a hybrid mutation
model that is a mixture of the stepwise mutation model and the infinite
alleles model. Both models are explained by Bruce Walsh in [13].

6.5 Modal Haplotype

If we have a group of people we are interested in the haplotype of their most
recent common ancestor. This is called the modal haplotype, base haplotype
or ancestral haplotype. The following table shows the haplotypes of three
persons on the 6-marker scale and their modal haplotype.

DYS393 DYS390 DYS19 DYS391 DYS385a DYS385b

modal 13 24 14 11 11 14
Peter 13 24 15 11 11 14
Carl 13 24 14 11 12 14
Clas 13 23 14 12 11 14

The modal haplotype can easily be calculated. If we use equation 4 to

calculate the genetic distance in years for just a single marker we see that it
takes 12500 years on average for a single marker to mutate only once. Thus
for most practical purposes a single marker rarely mutates at all. So for a
single marker we just look which value occurs most often and take this as
the modal value.
Phylofriend uses this simplified approach. For many thousands of years
it is no longer valid. Calculating the median values would be a better choice
then. The problem with the median approach is that some markers show
rather big leaps when mutating. This could lead to wrong results. I recom-
mend that you always take a good look at your data to see which method
would be appropriate.

6.6 Age Calculation

6.6.1 Mutation Counting
The simplest method to calculate the time to the most recent common an-
cestor (TMRCA) for a group of people is mutation counting as described in
great detail by Anatole Klyosov in [8]. Here a give a short overview only.
First we calculate the modal (ancestral) haplotype. Then we count the
mutations of every single group member relative to the modal haplotype and

20
take the average. This should be a good estimate for the group’s genetic
distance to their ancestor.
Again we use the definition of the mutationrate 1 but this time we use
the average value:
m̄
k= (6)
gy
k: Mutation rate per marker and generation
m̄: Average number of mutations relative to modal haplotype
g: Number of generations
y: Number of Y-STR values per haplotype

As mutations occur by coincidence equation 6 is only valid for sufficiently

high values of m̄. By using the definition of the average, equation 6 is equiv-
alent to
PN
mi
i=1
k= (7)
N gy
or
PN
mi
i=1
g= (8)
N ky
g: Number of generations
mi : Number of mutations of haplotype i relative to modal haplotype
N: Number of haplotypes (persons)
k: Mutation rate per marker and generation
y: Number of Y-STR values per haplotype

Equation 8 is the time to the most recent common ancestor in generations.

If we know the generation time d we can write this (see equation 3) as
PN
i=1mi
t=d (9)
N ky
t: Time to most recent common ancestor (TMRCA)
d: Generation time
mi : Number of mutations of haplotype i relative to modal haplotype
N: Number of haplotypes (persons)
k: Mutation rate per marker and generation
y: Number of Y-STR values per haplotype

21
 The following example shows how to use the formula but it is intended as
an easy to understand educational exercise. For real world applications more
haplotypes (rule of thumb: use at least 5) and a bigger number of average
mutations (4 or more is quite good) is needed.
The table on page 20 contains three haplotypes on the 6-marker scale
(N = 3, y = 6). Peter and Carl differ by one mutation to the modal hap-
lotype and Clas by two. We use a generation distance of 25 years (d = 25)
and the average mutation rate on the 6-marker scale is 0.002[8]. So we get
1+1+2
t = 25 · = 2778 years (10)
3 · 0.002 · 6
Of course the simple method of mutation counting comes not without
problems. The biggest one is the correct distribution of the population sam-
ple. For example if you want to calculate the TMRCA for a family which is
just a few hundred years old and add a person by accident who is at a distance
of several thousand years your TMRCA value will be much too large.
On the other hand if you want to calculate the age of a whole haplogroup
and add a large number of members from one family who are all very close
together your TMRCA value will be much too small. So this method must
be handled with care.
A more general problem is that mutation rates are derived from popu-
lation samples and there is no guarantee that they fit to your data set. So
whenever possible try to find some historical event to calibrate your data.

6.7 From Distances to Trees

Phylofriend computes genetic distances between different persons. These
distances can be used to derive phylogenetic trees using standard algorithms
that are implemented by the PHYLIP[6] program.
Genetic distances are often represented by two-dimensional drawings that
are easy to understand. Here we label the distance components as Component
1 and Component 2. Usually the values for each component are derived from
many markers by using principal component analysis[11].
Figure 4 shows the genetic positions of several persons and the corre-
sponding phylogenetic trees. In the first example Carl and Clas are close
together. So they are also grouped together in the tree. Peter is at a longer
distance but the distance to Carl and Clas is the same. Thus he is put into
a separate branch. The genetic distance between the persons is represented
by the length of the horizontal lines in the tree.
The second example is slightly more complicated. Carl and Clas are still
close together but Peter is at a different distance to each of them. The trees

22
Figure 4: Genetic distances can be represented by phylogenetic trees. Dif-
ferent algorithms often yield different results.

were constructed by using two different methods, Fitch-Margoliash (kitch

program) and Neighbor. Fitch-Margoliash groups Carl and Clas together
and puts Peter at a medium distance from both of them. This is a sensible
approach because some people develop many mutation and others develop
less. So the ancestor of Carl and Clas was probably somewere in-between.
However the resulting tree does not represent the genetic distances correctly
but it displays a very nice representation of temporal evolution.
The neighbor method tries to represent the genetic distances as good as
possible. In this case it is accurate but it gives you a bad idea of temporal
evolution.
For a high quality data set both methods should yield similar results
in grouping people together. The real problem is that mutations occur by
coincidence and we often have a very sparse collection of descendants who
are at large genetic distances to each other.
Figure 5 shows a more realistic situation. It represents the genetic po-
sitions of members from two different families. All members of one family
descent from one common ancestor. Due to the laws of statistics most family
members are close to his haplotype. So it can be calculated. It is called the
ancestral or modal haplotype.
Other family member are at a greater distance from the ancestor. Some-

23
Figure 5: Haplotypes and genetic distances of two families. Although both
lineages have a different modal (ancestral) haplotype the haplotypes of in-
dividuals from both families are sometimes close together. In such cases
phylogenetic tree algorithms yield wrong results.

times members from different families come close to each other. In such cases
all algorithms will give you wrong results.
This is the reason why it is so important to test for haplogroups before
creating a phylogenetic tree. Haplogroups are caused by very stable muta-
tions. Haplotypes often overlap. So it should be ensured that all persons in
a tree belong to the same haplogroup.

24
References
[1] Dmitry Adamov, Vladimir Guryanov, Sergey Karzhavin, Vladimir
Tagankin, Vadim Urasin. Defining a New Rate Constant for
Y-Chromosome SNPs based on Full Sequencing Data. The Russian
Journal of Genetic Genealogy (Ðóññêàÿ âåðñèÿ), Vol 6, No 2
(2014)/Vol 7, No 1 (2015).

[2] James Archie, William H.E. Day, Wayne Maddison, Christopher

Meacham, F. James Rohlf, David Swofford, Joe Felsenstein, The
Newick Tree Format. 1986, Date accessed: 2014-08-06.

[3] Alix Boc, Alpha Boubacar Diallo, Vladimir Makarenkov, T-REX: a

web server for inferring, validating and visualizing phylogenetic trees
and networks. Nucleic Acids Research (2012) 40 (W1): W573-W579.
doi: 10.1093/nar/gks485, 2012.

[4] R. A. Canada, How does the infinite allele comparison method work
for palindromic markers?. Family Tree DNA, 2014, Date accessed:
2014-08-07.

[5] Family Tree DNA, Big Y White Paper. Family Tree DNA, August
2014.

[6] Joe Felsenstein, PHYLIP, a free package of programs for inferring

phylogenies.. University of Washington, First version: 1980, Date
accessed: 2014-08-06.

[7] David Hamilton, An accurate genetic clock, bioRxiv preprint, first

posted online June 15, 2015, doi: 10.1101/020933.

[8] Anatole A. Klyosov, DNA Genealogy, Mutation Rates, and Some

Historical Evidence Written in Y-Chromosome, Part I: Basic
Principles and the Method. Journal of Genetic Genealogy,
5(2):186-216, 2009.

[9] Anatole A. Klyosov, Ancient History of the Arbins, Bearers of

Haplogroup R1b, from Central Asia to Europe, 16,000 to 1500 Years
before Present. Advances in Anthropology, Vol.2, No.2, 87-105,
doi:10.4236/aa.2012.22010, 2012.

[10] ISOGG, Listing Criteria for SNP Inclusion into the ISOGG Y-DNA
Haplogroup Tree - 2015. Date visited: 2015-10-06.

25
[11] Jonathon Shlens, A Tutorial on Principal Component Analysis.
Center for Neural Science, New York University, New York City,
Systems Neurobiology Laboratory, Salk Insitute for Biological Studies,
La Jolla, 2009.

[12] Wikipedia, List of Y-STR markers. Date accessed: 2014-08-15.

[13] Bruce Walsh, Details on the various assumptions used in computing

TMRCA. Family Tree DNA, 2002, Date accessed: 2014-08-07.

[14] Bruce Walsh, Estimating the Time to the Most Recent Common
Ancestor for the Y chromosome or Mitochondrial DNA for a Pair of
Individuals. Genetics Society of America, 2001.

26