›Introduction to DNA and Genealogy
  ››mtDNA and Y-DNA
 ›Meiosis and Crossover


Click here to help      
so Angela can improve these pages

**Page under construction - changes/edits will be made on a daily basis***

The Autosomal DNA contains most of the DNA that makes us who we are. Because of the way it is transferred down the generations, we usually cannot identify which specific ancestor gave us specific genes (unless we compare our results with people we know we are related to, and see which stretches of DNA match).

Below is a diagram showing how much of our autosomal DNA we inherit from our Ancestors. As you can see, our Autosomal DNA is (theoretically) representative of our whole ancestry.

Up until a couple of years ago, - the only autosomal tests that were available for comparing between people were those using CODIS markers (ie. the DNA markers used for Crime scene DNA testing). These use only a dozen or so markers, so were of limited usefulness. These markers were never intended to be used to infer biogeographical ancestry, - but there are some companies that claim to do so and they can give very misleading results. These CODIS markers are also the same markers that are commonly used for paternity testing. They can also be used for testing other relationship (eg. if two people are siblings), but often results from these markers can be inconclusive. For interest I had my family tested with these markers and the results can be seen here.

The latest new offering in the genetic genealogy world are products like 23andMe's "relative finder"and FTDNA's "Family finder". These test more than half a million DNA SNP markers so are considerably more robust, and therefore far more scientifically valid.


Grandparent/Child inheritance diagrams

These diagrams are constructed using the genetic data of a real family. They were constructed from the 23andMe "family inheritance" function. The genome of each "child" was compared to 3 of their 4 grandparents, and the comparison diagrams were spliced together to create the diagrams below.


As you will see from the diagrams above, each child inherits 50% of their DNA from each parent (ie. the genetic contribution from both paternal grandparents adds up to 50%, and the genetic contribution from both maternal grandparents adds up to 50%), but the contribution from each individual grandparent varies from the predicted 25%... Why is this so???

On average we inherit 25% from each grandparent, but in reality it can vary anywhere from 0%-50%.
To understand how this is possible, I will use the analogy of coin tossing. Think for a minute of coin tossing -
If a coin is tossed 50 times, the mathematical average result would be 25 heads and 25 tails. However, in reality the number of heads and tails varies. The more times you throw the coin, the closer you get to a 50:50 average, but the smaller the number of coin tosses, then the greater the variability of results.
In Mathematical probability theory there is something called a "normal curve", and according to this in about 90% of cases the probability of getting a head rather than a tail (or vice versa) ranges from approximately 40-60%

So,.. - how many DNA "coin tosses" are there??... We have 23 pairs of chromosomes, but there are there are billions of SNP's. So,.. is it 23 coin tosses? or is it billions of coin tosses?
...In fact - it is neither!. Strictly speaking it is 23 "coin tosses", - but before the "coin tossing" happens, something special happens to the pairs of chromosomes called "crossover". We can see the evidence of "crossover"in the diagrams above. Notice how in most cases each chromosome inherited from each parent is a blend of the 2 grandparents rather than from just one or the other grandparent?

You can read more about crossover (and Meiosis) here

Finding cousins with Autosomal DNA

At present there are two Autosomal DNA products available that are specifically designed to help with genealogy. In October 2009, 23andMe lauched the beta for "Relative finder", and in February 2010 FTDNA began its launch of "Family Finder". I will add more information about FTDNA's family finder as more information is known...

Theoretical %'s shared with different types of cousin

                                                                   x                                                x
                                                      x                                                x                                             x
                                         x                                                x                                             x                                                   x
                             x                                                x                                             x                                                  x                                                x
   x                                                             x                                             x                                                   x                                                x                                                    x

Vala and Jolinar share 2 of their 128 5x Great Grandparents

Jennifer and Jolinar share 2 of their 64 4x Great Grandparents Cameron and Jolinar share 2 of their 32 3x Great Grandparents Tamera and Jolinar share 2 of their 16 2x Great Grandparents Matthew and Jolinar share 2 of their 8 Great Grandparents Olivia and Jolinar share 2 of their 4 Grandparents Jacob, Daniel and Jolinar share all 4 of their 4 grandparents

As we saw just before,.. there is variation in the %'s of DNA from each grandparent that are passed on by parents to children. This variation would tend to compound through the generations, so thus there would in reality be a wide range of variation. On average sixth cousins would share .01% of their DNA, but in reality it could easily vary anywhere between 0% - 1%.

Autosomal DNA can be used to estimate how related two people are - but you can see that because of the randomness that happens when each persons DNA is halved to make the egg or sperm, there is a wide degree of variation, and this large degree of variation is compounded each generation. So thus, if the amount of matching DNA that two people have is consistent with their being 5th cousins, they may be in reality more related, or less related.

Therefore if two people have a match in a tool such as 23andMes relative finder or FTDNA's Family finder that suggests that they are 5th cousins, but the common link can't be found, - then we can't really blame the tool for "making a mistake", - in a way,... if anyone is to "blame" it is our ancestors, for passing down each of their parents DNA in such a random way. Likewise, if we compare our DNA with cousins that we have already identified with conventional genealogy, we may find that the estimated relationship (from the DNA) may vary a bit from the known relationship - it may overestimate, or it may underestimate. Again, we can't "blame" tools such as relative finder for not estimating the true relationship "correctly"- it's our ancestors that are really to "blame".

Accordingly,.. FTDNA has chosen to report only to 5th cousin level - as they know that as we go back each generation, the margin of error increases, and beyond the 5th cousin level the margin of error is too great to give a reliable result.

Comparison between different autosomal tests
(for Ancestry information)

What is tested? Half a million Autosomal SNP's 13876 X SNP's 176 "Geographically informative" autosomal markers 1 Million SNP's
15 - 27 CODIS autosomal markers
Half a million autosomal SNP's
Comparing close relatives (2nd cousins & closer)
Health traits
"Global similarity"
"Ancestry painting"
Value for Money


A comparison of my 23andMe results, DecodeMe results (from upload of 23andMe data), and DNA tribes can be seen here [link to be added soon].

If you are interested primarily in Genealogy - I recommend FTDNA's "Family finder". FTDNA is primarily a Genealogy company, so any matches you might have are more likely to also be genealogists. Many of 23andMe's customers are not genealogists, so the response rate from "relative finder" matches is consequently relatively low. FTDNA's chromosome browser is also the best for comparing matching segments - it combines the best features of both of its main competitors. Like Decodeme, you can select between several different thresholds for matching segment lengths. Like 23andMe, FTDNA's "Family finder" matches you with other customers in their database that have matching stetches of DNA (and thus may have inherited the stretch from a common ancestor).

If you are interested primarily in comparing close relatives - I recommend either 23andMe or FTDNA's "Family finder". If you want to compare siblings, then 23andMe would be the preferred choice (since FTDNA doesn't distinguish between half identical and fully identical chromosomal regions, and full siblings will have some fully identical segments). If you are not interested in sibling comparisons and not interested in family inheritance of health traits, then FTDNA's "family finder" would be better value for money.

However, If you are wanting a legal paternity or siblingship test, then you instead need to take a legal AABB accredited test (which usually use CODIS markers).
At present, results of tests such as 23andMe and FTDNA's "family finder" are not legally admissible in family law proceedings. If you want to test merely for peace of mind, and do not require the tests for legal purposes then autosomal tests such as 23andMe and FTDNA's family finder will (I suspect) in most instances give clearer and more definitive answers than CODIS autosomal markers can.If you need a legal test to determine paternity/maternity etc. then the best company to use will depend on what country you live in. To be admissible in legal proceedings, the "chain of custody" needs to be 100% - ie. the court needs to be 100% sure that the sample came from the person that the sample is alledged to be from. This means that the test is taken in the presence of a representative/agent of the DNA company who can therefore verify who the sample was taken from. For example, a person who lives in New Zealand can only test with a US company if they physically go to the US themselves. This would be impractical, so therefore they would need to use a New Zealand company, - they canot order a test from a US company, test in their own home and then send the sample to the DNA company - because the DNA company would not have proof that the test was taken by the person that the sample is alleged to be from, and it would therefore not be legally admissible.

If you are primarily interested in "Racial" makeup, - I recommend 23andMe. The results from 23andMe are more scientifically robust than the results from "Ancestry by DNA", and are less confusing than the results from Decodeme. People with just European ancestry generally are found to be 100% "European".

However: Some people of predominantly East African ancestry might find that the test says that they have some "European" and some "Asian" ancestry. People with Asiatic Indian ancestry also tend to find that their ancestry painting indicates that they have a higher percentage of "European" than they have "Asian".
This most likely reflects the fact that human "racial" genetic variation doesn't in reality "fit" into 3 discrete racial "Boxes" - instead there are in reality, gradients of variation. It also might reflect the fact that Europeans and Asians were most likely derived from East Africans, so some of the gene variants that are more typical of people of "Asian" or "European" ancestry may have originated in their original Eastern African ancestors. Likewise, the large "European" scores that Asiatic Indians have, might reflect that many of the "European" gene variants originated in the common ancestors of "Europeans" and Asiatic Indians, and perhaps a better label would be "Eurasian". Thus there will be some people that do not have admixed ancestries, yet they will appear as a mix of two or all three"racial" categories. This isn't because there is something wrong with the algorithm used - this merely reflects that variation between "racial" groups is variation along continuums, and all that the algorithms do is artificially "force" the gradual variation into discrete categories. Sometimes the underlying real biological variation does not correlate with racial identity and/or human social perception.

If you are primarily interested in Health traits - I recommend either DecodeMe or 23andMe. DecodeMe's coverage of some conditions is more robust, but overall 23andMe is better value for money. If you are only interested in health traits (and not interested in Ancestry information at all), then Navigenics is another option that you might like to consider.

If you are primarily interested in "Global similarity" - I recommend DecodeMe. 23andMe's advanced global similarity is excellent,.. but DecodeMe's is beyond excellent. Get Decode me if you can afford it,.. otherwise 23andMe is better value for money.

However: Both 23andMe and DecodeMe's "global similarity" results would benefit from additional reference populations. 23andMe places me in the "Irish" cluster, and DecodeMe places me between the french and Scandinavian clusters. My ancestry is approximately 65% Scottish, 33% English, 1% Welsh, 1% "Jamaican mix". 23andMe does not have Scotland as a reference population and Decode me does not have any British Isles reference populations (apart from Orkney - which would not be representative of the whole of the Bristish Isles).

If you are primarily interested in Entertainment, - I recommend DNA tribes. CODIS markers are the markers used in forensics (eg. CSI, paternity). The markers were selected for their ability to distinguish between individuals, not for their ability to predict biogeographical ancestry. Their application to predict ancestry is based solely on false logic, and is not scientifically valid.

As an example: DNA tribes say my highest geographical match is from India (99% of my ancestors come from the British Isles, and the last time any of my ancestors were anywhere near India was thousands of years ago). Based on my CODIS marker values, my combination of marker values is more likely to occur by chance alone, in India. When CODIS markers are used by forensic scientists, values are compared to the values found in various different populations to give an estimate of the chance of a false match. In other words, if the crime scene DNA matches a suspects DNA,- how confident can they be that the crime scene DNA came from the suspect? (v's the chance that the suspects DNA matches from chance alone).
It is scientific to say that in the context of the population of India, the chances of my DNA matching another persons exactly from chance alone, is marginally less remote in India compared to other geographical regions - which is not the same as saying my ancestory most likely came from India. It just says that I have a slightly higher chance of being falsely matched to a criminals crime scene DNA in India. Overall though, the chances of me randomly matching a criminals DNA are incredibly incredibly slim.


Creative Commons License
This work is licenced under a Creative Commons Licence.