PDA

View Full Version : Auto cluster



firemonkey
10-12-2019, 11:08 AM
Is there any way to adjust the settings for it?

msmarjoribanks
10-12-2019, 06:40 PM
It's based on the tool at GeneticAffairs, which does allow you to adjust the settings, but I don't see how to do it at MyHeritage.

firemonkey
10-12-2019, 09:46 PM
Unfortunately Genetic affairs doesn't accept profiles from My H . I was interested to see if I could get a cluster with Scandinavian matches . I used FTDNA.

Cluster 64 4 people

HA Scandinavian name
MU British name
BO Scandinavian name
LO British name

HA BO are both also on My H .

HA= just Scandinavian in common
BO= Eng ,Scand and Finn in common .

firemonkey
10-13-2019, 01:41 AM
Re uploaded with FTDNA .


Now HA BO LO RO form a cluster . However when using those with FTDNA matrix RO doesn't go with BO and LO. It seems to my uneducated(on this matter) mind that auto cluster(genetic affairs) and matrix(FTDNA) are not the same type of thing just with different names.


I used HA BO LO RO with FTDNA's chromosome browser

HA BO LO = chr 7
RO= chr9 30514250-68228343 which I believe is in a pileup region .

msmarjoribanks
10-13-2019, 06:12 PM
The settings may effect whether a particular person is assigned to a cluster or not. What I've found is that there are overlapping clusters, so it's useful to look at whether people in one cluster also have lots of matches with people in another.

Why do you think they are different?

This is what MyHeritage says: https://blog.myheritage.com/2019/02/introducing-autoclusters-for-dna-matches/

"This new tool was developed in collaboration with Evert-Jan Blom of GeneticAffairs.com, based on technology that he created, further enhanced by the MyHeritage team. Our enhancements include better clustering of endogamous populations (people who lived in isolated communities with a high rate of intermarriages, such as Ashkenazi Jews and Acadians), and automatic threshold selection for optimal clustering so that users need not experiment with any parameters."

So that sounds like you cannot change the thresholds.

firemonkey
10-13-2019, 06:59 PM
If I say I'm a little confused by auto clustering that's putting it mildly . I'm struggling to see how shared matches= a common ancestor when 3 matches are on chr7 , and the other in a pileup region on chr9 .

msmarjoribanks
10-13-2019, 08:17 PM
If I say I'm a little confused by auto clustering that's putting it mildly . I'm struggling to see how shared matches= a common ancestor when 3 matches are on chr7 , and the other in a pileup region on chr9 .

It doesn't triangulate. What it does is find a group of matches where you all share with each other. In theory, the one that shares with you on the pile-up region ought to be sharing with the other two. It's the fact it's a comparison group that allows the hypothesis that you are all descended from the same ancestors, although with only 4 that's not as confidence-inducing.

My MyHeritage matches seem to require a certain amount of shared cMs with the people in the cluster (presumably how they are protecting against the endogamy-related matches). What that does is exclude some likely clusters, but prevent ones just based on pileup regions. I don't find it helpful with FTDNA, because I don't get enough matches. It's helpful for me with Ancestry, but Ancestry has added enough other tools that allow one to do the same thing.

I'll look at my FTDNA clusters to see if they are pulling up likely pileup-only matches. (I have my parents on FTDNA, so mostly just look at their clusters, and haven't looked at those for a while.)

JMcB
10-13-2019, 09:54 PM
The settings may effect whether a particular person is assigned to a cluster or not. What I've found is that there are overlapping clusters, so it's useful to look at whether people in one cluster also have lots of matches with people in another.

Why do you think they are different?

This is what MyHeritage says: https://blog.myheritage.com/2019/02/introducing-autoclusters-for-dna-matches/

"This new tool was developed in collaboration with Evert-Jan Blom of GeneticAffairs.com, based on technology that he created, further enhanced by the MyHeritage team. Our enhancements include better clustering of endogamous populations (people who lived in isolated communities with a high rate of intermarriages, such as Ashkenazi Jews and Acadians), and automatic threshold selection for optimal clustering so that users need not experiment with any parameters."

So that sounds like you cannot change the thresholds.

Interesting. I was just thinking about uploading my tree to My Heritage, after doing a free raw data upload many months ago. Mostly I’ve ignored the site ever since because their results weren’t all that good. Do you happen to know what their Premium level goes for?

msmarjoribanks
10-13-2019, 11:42 PM
Interesting. I was just thinking about uploading my tree to My Heritage, after doing a free raw data upload many months ago. Mostly I’ve ignored the site ever since because their results weren’t all that good. Do you happen to know what their Premium level goes for?

I believe it's $99 for the first year and $129 after that. I think you can upload a rudimentary tree for free. I've never tested there, but it appears they are currently having a $49 special, although that probably comes with no subscription advantages.

I like the site despite the weird autosomal results since I could upload my parents and sister there (my sister is on Ancestry, my parents on FTDNA, I'm on both) and can sort matches that way (helps weed out false positives).

This discussion inspired me to check my mom's FTDNA and my own 23andMe and MyHeritage clusters. I must have my mom on a different (smaller) setting, as she gets 76 FTDNA clusters, many 2-3 people, whereas I get 7 clusters on 23andMe (I can recognize pretty much all of them, although some overlap), and about 16 on MH (I can recognize about 6 of the 16). My mom gets multiple all Swedish clusters at FTDNA, but she has a Swedish grandmother so that's not unexpected.

I am focused now on researching one specific branch (on my mother's side), and although I have a few matches on FTDNA and, especially, MH and 23andMe, the clusters are unhelpful for that branch. The vast majority of my relevant matches are on Ancestry. (I have way more clusters there, and my sister and my clusters are broken up differently, so we get a different mix of relevant people. Because Ancestry tends to underreport the amount of the relationship, at least for many of my cousins, it's annoying in a different way.)