Saturday, May 20, 2017

Lazarus Issues

[Note: As of May 25, 2017, GedMatch has a beta fix to Lazarus which will hopefully be pushed out shortly.]

Lazarus is one of the Tier 1 utilities offered by GedMatch.  It allows you to recreate at least some of an ancestor's genome, based on living relatives who have tested.  It takes two sets of kits--one set consists of descendants of the ancestor whose DNA you're trying to recreate (Group 1); the second set consists of non-descendant relatives of that individual (Group 2).

The utility basically looks at all of the DNA of those Group 1 descendants--some of which is that of the ancestor of interest but some of which is also those of other ancestors--and what segments those descendants have in common with known cousins, nieces, nephews, etc. from Group 2.  Those common segments should all come from the ancestor whose genome is being recreated.  (There is always the possibility, especially in endogamous populations, that you're including segments that are not inherited from that ancestor, but you can also include segments only over a certain size to try to account for that--but that's not the issue being raised in this post.)

I've created Lazarus kits of several of my ancestors.  But recently I've noticed that GedMatch no longer does comparisons of Lazarus kits to others correctly.  As an example, I recreated the genome of my great grandfather, Avraham Tzvi Diamond.  I've included five of his direct descendants (Group 1) and three relatives who are not descendants (Group 2)--a great niece, a great-great nephew, and a great-great niece.
Chart showing relationships.  Group 1 individuals are blue, Group 2 are green, and the kit being Lazarused is purple



When I create Avraham Tzvi's genome, there will obviously be some gaps, since likely some of his DNA was not inherited by these five in Group 1, and there is other DNA that may be inherited by some in Group 1 but not of anyone in Group 2, so it can't be attributed to him.  But using these two groups, Lazarus tells me that it can create a kit containing 1147.7cM of the actual 6800cMs (looking at both sets of chromosomes) that were in his DNA.  But even with this small number, I can still run comparisons between this Lazarus kit and other individuals to find some shared segments.

If I run Avraham Tzvi Diamond's Lazarus kit against Beth, I should get some subset of the DNA that she actually has in common with him.  But here's the issue--I'm getting way more than I should.

If I look at Beth's kit and the DNA that she shares with the individuals from Group 1, I get what's in the image below.  She should match no more than these segments in Avraham Tzvi's kit (and probably significantly less, since this displays any segment she has in common with Group 1 individuals of 1cM or greater, while the Lazarus kit only looked at segments of 7cM or greater).
DNA shared by Beth with Group 1 Individuals

However, a One-To-One comparison of Beth's kit with Avraham Tzvi's Lazarus kit gives me significantly more shared DNA than we should know she shares with him.
DNA that GedMatch says Beth shares with Avraham Tzvi's Lazarus Kit

For example, look at Chromosome 1 (what better place to start?).  On the image showing DNA which Beth shares with individuals in Group 1, she has some decent-sized segments to the left of the centomere--although with gaps between those segments--but has very few segments to the right (and all of those are less than 7cM in length.  But the One-To-One between Beth and Avraham Tzvi's Lazarus kit had 3 large, almost contiguous segments in common that essentially span nearly the entire chromosome.  This makes absolutely no sense.  This analysis can be done for all of Beth's chromosomes; she is said to share significantly more DNA with Avraham Tzvi than she possibly can, with the information that we know.

In fact, if you look at the "Resulting Segments" which is displayed when creating this Lazarus kit, Beth is matching segments that aren't in this kit--which shouldn't happen.
Segments in Avraham Tzvi's Lazarused Kit (which don't include segments Beth is said to match with him)

This issue holds true for all of those in Group 2.  And it seems to hold true for all of my Lazarus kits.  And if you look at Beth's contributions when Lazarus is run, it is lacking all of those large segments that GedMatch says she shares with that Lazarus kit, which shouldn't be the case.

It looks like Lazarus is processing correctly, but it looks like GedMatch isn't calculating Lazarus One-To-One comparisons correctly.

Has anyone else seen similar issues?  Any idea of what's happening?  Please comment below.

Note:  I'm on Twitter.  Follow me (@larasgenealogy).

Want to get future blog posts emailed to you automatically?
Enter your email address:

5 comments:

  1. OK, so I tried this using a very simple kit. I put myself alone in Group 1 and a third cousin and a 3C1R in Group 2.

    The "Contributions" table showed nine segments (excluding chromosome 23). Then I did a one-on-one with me and the Lazarus kit. Everything came out right except one chromosome.

    "Contributions" shows two segments 8,963,653-59,890,187 (51 cM) and 82,425,667-90,201,144 (7.1 cM). That's a total of 58.1 cM.

    The one-to-one shows that chromosome as 8,966,210-90,188,979 (83.4 cM).

    This is obviously wrong.

    ReplyDelete
    Replies
    1. I just did a one-on-one with the person in Group 2 with the problematic chromosome - he also shows 83.4 cM.

      Delete
  2. Adding Aunt Betty to Group 1 runs MY total on the problem chromosome to 108.8, but what's really odd is that it extends my match to 115,834,939 which is where A Betty's match with the other Group 2 member begins.

    ReplyDelete
  3. That's no good! Here's hoping Gedmatch is able to fix the problem.

    Prior to this discovery, had you ever used this tool to recreate the genome for more distant ancestors? I've been wondering if this would work for a 5th great-grandfather or if that's just too distant with autosomal DNA.

    ReplyDelete
    Replies
    1. I've recreated for ancestors as far back as great-great-great grandparents. The problem with further is you need relatives of that person who aren't descendants. And particularly when you're endogamous, any DNA you share with those 6th or so cousins may not even be from that common ancestor, so the kit isn't useful. I usually run Lazarus looking at reasonably large segments anyways (but one I realized things were weird, for this analysis I used their default values).

      Delete