Kurd

12-02-2018, 02:53 PM

Users frequently ask me why their % of a component varies so much from calculator to calculator and would like to know whether that indicates that the calculator is flawed. For example a K3 calculator with components; W Eurasian, E Eurasian, African will show them with 10% E Eurasian, whereas a K30 calculator will show them as only 1% E Eurasian.

I decided to take a little time to answer this since it seems like most would like to know.

Actually this does not necessarily indicate a flaw with the calculator. In fact the problem is not even related to genetics, it has to do with how a mathematical fraction is calculated.

Here is a simple example:

Let's assume we have 3 different K3 calculators. 2 of the 3 components are fixed in all 3 calculators, the 3rd is allowed to vary. So say:

- Calculator 1 (C1): Components are W European, E Asian, African

- Calculator 2 (C2): Components are Neanderthal, E Asian, African

- Calculator 3 (C3): Components are Your close relatives, E Asian, African

To further simplify, lets say you are a S Asian and the calculator calculates that you share this many alleles with the various components in the above 3 calculators:

E Asian

African

W European

Neanderthal

Your close relatives

Number of alleles you share with

20

3

5

0

1000

Now let's calculate your E Asian % for the 3 calculators:

For C1 we have E Asian = 20 / (20+3+5) = 71%

For C2 we have E Asian = 20 / (0+20+3) = 87%

For C3 we have E Asian = 20 / (1000+20+3) = 1.9%

So which one is correct?

The answer is they all are, and that the admixture % will depend what the OTHER calculator components are. This is the reason in my next generation calculators which I call SAPDA ( a program for inference of Shared Ancestral Population Defining Alleles) for which I have been refining the script over the past 6 months, the user gets a graphical output of the number of population defining mutations they share with the calculator's source populations, in addition to the classical admixture pie-charts. This way the user has a 1 to 1 comparison with the various population sources in the calculator.

The outlet for this calculator is still undecided, but announcements will be made.

Recognizing that knowledge is power and most users would like more than just a classical admixture pie-chart, I have written the software so that the user gets a total of 11 PDF graphs and charts about various aspects of their genome

I decided to take a little time to answer this since it seems like most would like to know.

Actually this does not necessarily indicate a flaw with the calculator. In fact the problem is not even related to genetics, it has to do with how a mathematical fraction is calculated.

Here is a simple example:

Let's assume we have 3 different K3 calculators. 2 of the 3 components are fixed in all 3 calculators, the 3rd is allowed to vary. So say:

- Calculator 1 (C1): Components are W European, E Asian, African

- Calculator 2 (C2): Components are Neanderthal, E Asian, African

- Calculator 3 (C3): Components are Your close relatives, E Asian, African

To further simplify, lets say you are a S Asian and the calculator calculates that you share this many alleles with the various components in the above 3 calculators:

E Asian

African

W European

Neanderthal

Your close relatives

Number of alleles you share with

20

3

5

0

1000

Now let's calculate your E Asian % for the 3 calculators:

For C1 we have E Asian = 20 / (20+3+5) = 71%

For C2 we have E Asian = 20 / (0+20+3) = 87%

For C3 we have E Asian = 20 / (1000+20+3) = 1.9%

So which one is correct?

The answer is they all are, and that the admixture % will depend what the OTHER calculator components are. This is the reason in my next generation calculators which I call SAPDA ( a program for inference of Shared Ancestral Population Defining Alleles) for which I have been refining the script over the past 6 months, the user gets a graphical output of the number of population defining mutations they share with the calculator's source populations, in addition to the classical admixture pie-charts. This way the user has a 1 to 1 comparison with the various population sources in the calculator.

The outlet for this calculator is still undecided, but announcements will be made.

Recognizing that knowledge is power and most users would like more than just a classical admixture pie-chart, I have written the software so that the user gets a total of 11 PDF graphs and charts about various aspects of their genome