DNA testing

Seanachai

When DNA testing is carried out is there a lot of gaps that are filled with speculation? It seems that the researchers are cross-referencing against groups that have already been tested, but does this give an accurate picture. Could they be missing out on markers that are staring them in the face because they don't really know what they are looking for?

Black Swan

Although a bit dated, Olga Troyanskaya, et al (2001) in Missing value estimation methods for DNA microarrays, Bioinformatics, Volume 17 Issue 6 Pp. 520-525, discussed a comparative study of several methods for the estimation of missing values.

dubhthach

Seanachai wrote: »

When DNA testing is carried out is there a lot of gaps that are filled with speculation? It seems that the researchers are cross-referencing against groups that have already been tested, but does this give an accurate picture. Could they be missing out on markers that are staring them in the face because they don't really know what they are looking for?

There are a number of projects out there that specifically aim at testing less known populations. There is of course issue of "Calculator effect" where if you only have a small number of reference populations the software will try and get a sane comparison between your sample and these references. Obviously if more reference populations are added to the comparison you could end up with slightly different result.

The newest thing of course is using ancient DNA to provide baseline to compare modern populations with. I see FTDNA have now added an option when looking at their FamilyFinder results that compares a sample against ancient ones. Here's mine:



With modern population samples I get the following in comparison (in AncestryDNA):


Unsurprising as 7 of my 8 great grandparents were born in Ireland (the 8th was "Liverpool Irish")

dubhthach

Leaving that aside new markers and variations are found everyday as more and more full genomes are published. One has to remember the first full human genome was published only in 2003 (give or take a year) and cost about $2.7billion, now adays you can get "medical grade" genomes done for about $1,000-$1,500 as a result there's massive explosion in data. That and we now have something like 200 published ancient DNA genomes from Palaeolithic right up to the Iron age/Anglo-Saxon period.

Seanachai

dubhthach wrote: »

Leaving that aside new markers and variations are found everyday as more and more full genomes are published. One has to remember the first full human genome was published only in 2003 (give or take a year) and cost about $2.7billion, now adays you can get "medical grade" genomes done for about $1,000-$1,500 as a result there's massive explosion in data. That and we now have something like 200 published ancient DNA genomes from Palaeolithic right up to the Iron age/Anglo-Saxon period.

I'm curious to know if Berber or other North African DNA shows up much in people with Irish surnames. I'll send in a sample next year.

dubhthach

Seanachai wrote: »

I'm curious to know if Berber or other North African DNA shows up much in people with Irish surnames. I'll send in a sample next year.

None that I've seen anyway's, but hardly surprising as there's no evidence for North African admixture in Northern Europe, now in Iberia that's a different story.

I submitted my own sample to dna.land who have a number of North African sample populations, came back as 0% North African in their analysis. Their samples populations from region been:

• Tunisian in Tunisia
• Algerian in Algeria
• Mozabite in Algeria
• Sahrawi in Western Sahara

Irish like most European's can be modelled as a three way admixture between:
1. Mesolithic like population
2. Neolithic European farmers
3. Metal age Steppe input

Seanachai

dubhthach wrote: »

None that I've seen anyway's, but hardly surprising as there's no evidence for North African admixture in Northern Europe, now in Iberia that's a different story.

I submitted my own sample to dna.land who have a number of North African sample populations, came back as 0% North African in their analysis. Their samples populations from region been:

• Tunisian in Tunisia
• Algerian in Algeria
• Mozabite in Algeria
• Sahrawi in Western Sahara

Irish like most European's can be modelled as a three way admixture between:
1. Mesolithic like population
2. Neolithic European farmers
3. Metal age Steppe input

Am I right in thinking that if a lot more people submitted samples then they could pinpoint the results more?

dubhthach

Seanachai wrote: »

Am I right in thinking that if a lot more people submitted samples then they could pinpoint the results more?

Well it's not necessary the amount of people but where the people are from. So if you only have a calculator with 5 sample populations (you could have 10,000 people in each sample) than when you run a individual through it you will only see how that person compares to those 5 samples.

So simplistic scenario above, imagine 5 samples are:
1. European's (using a sample set made up only of Tuscan's)
2. African (using a sample set purely from Nigeria)
3. Asian (using sample set purely from China)
4. Native American (using sample set drawn from Peru)
5. Oceanic (Using sample set of Maori's)

In that scenario above if you tested the software calculator would conclude that you are closest to modern Tuscan's (Europeans in above scenario), which isn't of much use.

Now if instead you had sample populations from every country in Europe, the end result would be considerably more fine-grained.

AncestryDNA for example use to have a "Southern European" component in their calculator, as they increased their reference populations they spilt that into two namely:

1. Iberian
2. Italy/Greece

In future update they'll add in a distinct Sardinian component.

Likewise they use to have Irish and British as one component, but after building a Irish sample reference they were able to spilt into two separate components eg. "Irish" and "British"

In couple more years (once the Irish DNA atlas is published etc.) they'll probably be able to say that you are x% Connacht, y% Airgiallian, z% Meath etc.

Leaving that aside Ancestry's DNA database will probably hit 3 million samples within the next 6-12 months.