RootsChat.Com
General => Ancestral Family Tree DNA Testing => Topic started by: davidft on Thursday 28 February 19 14:33 GMT (UK)
-
MyHeritage has introduced a new Auto Clustering tool
"AutoClusters
An automatic tool that organizes your DNA Matches into clusters that likely descended from common ancestors.
AutoClustering organizes your MyHeritage DNA Matches into shared match clusters that likely descended from common ancestors. Each of the colored cells represents an intersection between two of your matches, meaning that both individuals match you and each other.
These cells are grouped together physically and by color to create a powerful visual chart of your shared match clusters. Each color represents one shared match cluster. Members of a cluster match you and most or all of the other cluster members."
It produces a report for you. Tells you which members it has excluded because they have too little information. It then groups your clusters which is helpful as it picks up ones you have missed and so acts as a prompt which members to follow if pursuing a particular line.
Yes this is very similar to the Auto clustering tool that was advertised on here last year except this is the company holding your results doing the clustering not some third party who then wants to charge you to access them. Previous thread on that linked below
https://www.rootschat.com/forum/index.php?topic=804727.36
-
Thanks, David. It is now autoclustering for me and could take a few hours, so I'll be patient.
It does look very like the Leeds method and what I'e been doing manually.
Gadget
-
I've just noticed that they also have a 'Theory of Family Relativity' which seems to list matches that share surnames and places and gives a tree as in Ancestry's Thrulines.
Gadget
-
Auto clustering sounds like a posh name for a traffic jam.
Martin
-
But less mobile.
It's taking it's time - not come through yet.
-
Sounds fantastic, but I can't see how to access this. Is it on the first page? I can only see the Theory of Relativity that Gadget referred to. This looks very much like the matching that ancestry has just brought out, that is, looking at various trees to see if they can work out where you match with someone. The first three suggestions for me they got exactly right, the fourth one way off the mark.
But it has the potential to be a good new tool if used correctly, like the ancestry one.
The companies are certainly upping their game, I bet myheritageDNA wishes they'd brought it out a day earlier!
Regards Margaret
Modified
Found it. DNA results>Tools
Auto clustering started for me and my husband.
-
Sounds fantastic, but I can't see how to access this. Is it on the first page?
Go to the DNA tab at the top of the page.
From the drop down menu select DNA Tools
It is then on the page that opens up
Good luck
-
Thanks David, Saw it and modified my message, then saw your reply.
Regards Margaret
-
Just seen this at the bottom of the auto clustering explanation page.
'The AutoClustering feature on MyHeritage was developed in collaboration with Evert-Jan Blom.'
Evert-Jan Blom was the chap who set up Genetic Affairs and who responded on the thread that David has linked to.
Regards Margaret
-
Just seen this at the bottom of the auto clustering explanation page.
'The AutoClustering feature on MyHeritage was developed in collaboration with Evert-Jan Blom.'
Evert-Jan Blom was the chap who set up Genetic Affairs and who responded on the thread that David has linked to.
Regards Margaret
Interesting. I saw the name but didn't connect it to the previous incarnation of the idea. Still feel better that the auto clustering is being done by the site and not a third party and Evert will have been recompensed for his idea I am sure so all is good. :)
-
i'm waiting for my clustering to come. Think it's taking awhile because loads using it. The figures I seem to think are fixed unlike genetic affairs where you can manipulate the values.I'll be interested to see the result of this one.
jean
-
I got all excited, reading this thread, as MyH did my test, but I don't subscribe.
Update, 18:42, David, could you comment, your original suggests that this free?
Martin
-
Just got my auto cluster report, but a bit confused about the accompanying explanatory pdf.
There is a section starting
The following matches have been excluded from the AutoCluster 19 analysis
because they did not have any shared matches:
and a section starting:
The following 130 matches met the inclusion criteria but ended up in singleton
clusters without other members and are therefore excluded from the analysis as well:
What's the difference between someone sharing no matches and being a member of a singleton cluster.
-
I received the folder at about 11 pm last might. I'm definitely underwhelmed ::)
My closest relative on there is a first cousin, once removed. She was in a group of her own although she matches with some of the 2nd cousins that we share.
-
LizzieL, I've got the same, 30 no shared matches, 139 singleton clusters. There must be an explanation but I can't think of one.
Gadget, Close relatives are excluded as they would be on too many clusters.
Someone asked previously, can't remember who, and you can adjust the settings, standard are matches between 20 cM and 350 cM, so probably wouldn't include your cousin once removed.
Mine apparently came through at 7 pm yesterday, but didn't see it yesterday, have only just turned laptop on, trying to see the results.
Having trouble extracting my zip file.
Regards Margaret
-
Like Gadget, my results took several hours to arrive. Looking at them this morning they have a lot more matches and also a lot more clusters, each with more members, than the ancestry set I did with GA using the default settings, so perhaps the default settings differ between sites or perhaps more of my distant cousins have their results on MH than Ancestry.
I like the animation - sad I know - puts me in mind of a very old computer game called battle chess.
It remains to be seen whether it enables any breakthroughs. I had worked out for myself that I probably have an NPE on one particular line three or four generations back because I have absolutely no DNA matches - not on Ancestry, not on FTDNA, not on MH, not on GEDMatch - to anyone who traces the same line back, but a reasonable scattering across almost all other lines. Perhaps this clustering will help me identify a candidate father.
Jane :-)
-
Gadget, Close relatives are excluded as they would be on too many clusters.
Margaret, it said this:
# The following 220 matches met the inclusion parameters but were placed in a cluster without other members and so are not included
She is in this list but my other close relatives are in the same cluster.
What I share with the top 3 - all related : 226/9; 215/12; 200/8 (ADDED - She is the 226 one)
I assume that the clustering is not just the cMs but also the segment matches,etc.
-
Jane
Ancestry has thrown up most of my close matches, although I do prefer My Heritage's tools.
Gadget
-
This sounds interesting. Now, the question is, will I be able to work out how to use it and will I understand the results? ;D
-
One day in and I have decided after playing around with the new auto clustering that I like it and it does have potential. OK yes I know it has more holes than a string vest but that is also true of Ancestry’s Trulines if what people have said about it in the other thread are anything to go by. I think what we all need to do is take a step back, accept it is new and hope that with time it will improve.
What I like about it
It gives you three reports, a PDF narrative, a HMTL document of the clusters and an excel document of the clusters. In my opinion the PDF narrative should be read first as this gives a quick summary telling you the minimum thresholds for inclusion in the reports, how many of your matches ended up in how many clusters e.g. 87 matches in 24 clusters. It then goes on to say how many of your matches have been excluded because they did not have any shared matches (i.e. I think that means you can not triangulate), and then finally it tells you how many matches ended up in clusters of one (201 in one case), not sure what this means other than they require more people to test to make these matches meaningful.
Whilst the clusters are displayed in colourful boxes do not ignore the little grey boxes they have a meaning as explained in the PDF and may help with particularly close matches.
The HTML document is where you get the colourful diagram (as shown in the opening post). This is not just a pretty picture! You can read it, and it does help explain where your matches match each other in a quick handy visual aid. This is then followed by a list of each clusters information telling you who is in what cluster, assign their tree a number so you can link to it, as well as showing who does not have a tree. It gives data on the match e.g. size, number of segments etc. Then in the final column “notes” it lists what you have done e.g. emailed the member, confirmed a match etc
The Excel document presents the information from the HTML document in excel format.
Making it work
I know some people have been disappointed how few clusters they have and that the theories of relativity are sometimes highly imaginative. First to say the system will only work if people put effort into it and to that extent the fuller your tree the more likely you are to click onto someone else’s tree. That is why it is a good idea to do your tree horizontally including siblings, cousins etc and not just vertically i.e. just direct ancestors. (That said the tree I Have on MyHeritage is just direct ancestors ;) )
It also helps if you have a near relative tested as this helps in forming the triangulated matches. I have both mine and my father’s data on there and this has helped make and confirm matches for me.
Then of course it requires lots of other people to test and be interested in following through with the project and the lack of this is what is hitting many people hard.
-
I have 29 clusters comprised of between 6 and 3 matches. Many of my matches are low (with only 12 above 50cM ), although over 300 above 20cM. The problem is that few of them have decent trees.
These matches are all from lines that I've thoroughly researched so, as yet, they haven't added very much to my knowledge of my ancestors.
I'm having problems with Thrulines and am awaiting a reply.
-
Bit confused by the exclusions.
My 3rd & 4th top matches, at 1 match of 62 cM for both, would appear to be a couple, either 2 sisters or a mother and daughter, as though they have different surnames their dna match to me is virtually identical, and of course they match each other too, ie it is the same dna segment that they share with me. In addition, one of the accounts is managed by the other.
Yet they are listed among the 118 excluded from the analysis because they are 'singleton clusters'.
Perhaps by definition, a singleton cluster is actually 2 people ( ie you can't be a cluster of 1).
Annoying, as they are the only matches in my top 6, that I don't really have any clues about!
My clusters are mainly made up of matches with very little dna correlation, to me, and I suspect, the progeny of ancestors who emigrated to the USA & Canada, or indeed Australia & New Zealand.
From experience, very nice people that they may be, but, like me they know little or nothing about where (in Ireland), their ancestors came from to Scotland, before emigrating further afield!!
The others in my top 6 are all from roughly the same branch of my family tree, with strong roots in the area that I live in, though only one of which I knew about before 'doing' my DNA!
And of course most of those in clusters have minimal family trees, mainly made up of 'Private'!
However, perhaps it i process the details methodically ( hard for me!), I may find useful pointers - I live in hope.
Jane
-
I have 29 clusters comprised of between 6 and 3 matches. Many of my matches are low (with only 12 above 50cM ), although over 300 above 20cM. The problem is that few of them have decent trees.
These matches are all from lines that I've thoroughly researched so, as yet, they haven't added very much to my knowledge of my ancestors.
I'm having problems with Thrulines and am awaiting a reply.
Just to add I am probably coming at this from a different standpoint from many in that I am not looking at it to extend my tree. I am using it to ascertain which parts of which chromosomes I get from which ancestors.
-
Just to add I am probably coming at this from a different standpoint from many in that I am not looking at it to extend my tree. I am using it to ascertain which parts of which chromosomes I get from which ancestors.
But, with a bit of work on the Chromosome browser, you can work it out yourself - and it's much more satisfying. A cousin and I spent a fun weekend a while ago doing just that!
-
David - an honest question -
Why do you want to know which particular segments of particular chromosomes you have inherited from particular ancestors? Is it curiosity, physical attributes, health, or ?
Gadget
-
is this a subscription only service? I can't access it even though MyHeritage performed my test.
Martin
-
David - an honest question -
Why do you want to know which particular segments of particular chromosomes you have inherited from particular ancestors? Is it curiosity, physical attributes, health, or ?
Gadget
Ah a question I can answer! Yes its purely curiosity and makes a nice change from searching through parish registers, at least for me.
-
is this a subscription only service? I can't access it even though MyHeritage performed my test.
Martin
I don't know for certain but I would strongly suspect you need a current subscription. I recall when I loaded my data to the site there were only certain things that you could access without a subscription. Others may know otherwise.
-
is this a subscription only service? I can't access it even though MyHeritage performed my test.
Martin
I don't subscribe, Martin, and I've got it. Are you looking in the right place?
DNA results page
Overview
Ethnicity estimate
DNA matches
Tools
Regards Margaret
-
I don't have a subscription, and I have accessed the clusters dna tool.
I have not seen the theory of family relativity, however! - co some advice re that would be good!!
-
Here's a thread, Jane
https://www.rootschat.com/forum/index.php?topic=809085.0
Gadget
Added - it's here:
https://www.rootschat.com/forum/index.php?topic=809085.msg6687970#msg6687970
-
In some respects MH is playing catch up. Collins Leeds Method has been working for a few months with DNAGedcom Client and Ancestry matches/icw. It also works with FTDNA & 23andMe and is in the process of extending to other testing companies and Genesis. There is also the DNAGedcom Chrome extension in conjunction with Genetic.Family where the hope is to be able to gather matches & ICWs from wherever you have tested and merge then cluster. It's a fast changing field...CLM works well for me with possibly a smattering of distant cousins.
-
Having had some time to look at MH clusters, I think they will be useful, but limited by the relatively narrow band of cMs they've chosen. This may well be because of the size of files generated by using wider limits, say, 20-400 cMs, may be just too great to email. I may be wrong? Significant matches do occur even below 15, although possibly beyond the scope of most family trees, without which detailed info we're struggling.
I'll be interested to see how MH develops this tool...
-
Yes the 25cM lower limit is restricting. For example, I've just discovered a 6th cousin (17.5 cM across 1 segment) using non-Beta Ancestry and trees rather than shared matching.
This particular line is very dominant in my matches. I think an interest in genealogy might be genetic
;D ;D ;D
Gadget
-
my second autocluster from my heritage took 25 hours to come but must admit it brought a couple of things to attention as to where to look for a match. It must be really backlogged the one from geetic affairs comes in less than 20 mins. On the whole though, doing an autocluster on ancestry because of the size in the database, there's a lot more matches in the cluster. So I think they still need to up their dna database on my heritage. I think they were using something like 350 and 50 with my heritage clusters whereas you can vary on genetic affairs but they may give the option eventually to manipulate the figures.
-
. I think they were using something like 350 and 50 with my heritage clusters
Mine took about 8 hours and the limits are 25cM and 350 cm.
Gadget
-
Interesting tool
Is it mindblowing? No.
But it helps the lazy ( like me) link up those odd names that make you scratch your head. Most of the names I had already linked in notes as being connected through trees and the 'match' section. But its nice to have confirmation
I may look to generate a report every 3 months and see the additions
-
Bit late to the thread but rather than raising a new one perhaps someone could help clarify something for me.
While I found it useful and in fact have requested it a second time and found a number of earlier cluster matches being discarded (I wanted to see if a more recent match appeared in any clusters), I'm somewhat confused as to how they can put a match in an auto cluster where there is no triangulated segment with him in the cluster. The cluster is 4 plus obviously me, I and only one other person in the cluster match this person at all and there is no triangulation between the three of us so how can this he be in this auto cluster? I match all of them on Chromosome 19, three of them on the same segment but not this other person who matches me on a different segment of the chromosome. On both auto cluster results this person has been put in this segment with the suggestion that there is a mutual ancestor, I can't see how they can come to that conclusion.
If anyone can help my brain cells understand this I'd be very grateful. ;D
-
I have also downloaded my clusters for a second time and they appear much different than the previous one a month ago. First time I had 42 clusters, the largest with 10 people, second batch was just 16 clusters, the largest being 5 members. Also when doing the triangulation only once have I found that 3 persons in a group, plus myself do actually triangulate. I too am somewhat confused.
-
Hi pollycat76, glad its not just me being confused with the auto-clusters on MyHeritage. :-\
-
Based on reading these messages, I decided to try this out. I don't have a subscription to My Heritage, but I did upload my Ancestry DNA and then paid for the 'unlock'. This seems to entitle me to use the DNA Tools.
I got the results back very quickly (my tree is only about 4,500 persons) and I already see some names that have shown up as having Common Ancestors on Ancestry. So let's see where this will lead!
Ms_C
-
Hi, I have also found that if you hover over the small grey coloured squares on the same line as the appropriate cluster, you will see two names appear say, John + Ben, (Ben is not part of John's cluster) If you follow the grey line up you will see that Ben appears in a separate cluster. You may have more than one grey area, showing different possible connections for the same person, which is quite helpful if you already "know" of this connection.
-
For those of you who are interested in clustering take a look at Genesis Gedmatch too if you have uploaded your results there. They have recently introduced a clustering function.