GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2015-08 > 1439670852


From: Ann Turner <>
Subject: Re: [DNA] GEDMatch 1:1 Thresholds?
Date: Sat, 15 Aug 2015 13:34:12 -0700
References: <735736465.1368544.1439656733029.JavaMail.yahoo@mail.yahoo.com>
In-Reply-To: <735736465.1368544.1439656733029.JavaMail.yahoo@mail.yahoo.com>


I don't know the answer to #1, unless somehow you had changed some of the
other parameters.

The cM segment length is more important than the number of SNPs. First of
all, the SNP count can vary over different regions of the genome. Also, the
SNP count can vary depending on which testing platform was used. If you are
comparing two 23andMe v3 kits, you'll automatically get higher SNP counts,
since there are close to a million SNPs. If you are comparing a 23andMe v4
kit with a Family Finder kit, the number of SNPs will be reduced, since
there are fewer SNPs in common between the two platforms. IMO, it's
appropriate to reduce the SNP threshold in that case, as long as you keep
the cM length high.

GEDmatch will generate a lot of false positive matches down at the 5 cM
level. See some examples here:

http://isogg.org/wiki/IBD#False_positive_matches

I think this is partly because GEDmatch doesn't have access to some of the
secondary requirements that FTDNA and 23andMe impose. For example, they
don't even report 5 cM matches unless there is also a longer segment (7+
cM). GEDmatch also allows more mismatches within a segment, so that's
something to watch out for (red lines in the one-to-one tool if you display
the graphic).

Ann Turner



On Sat, Aug 15, 2015 at 9:38 AM, Wesley Johnston via <
> wrote:

> I compared two pairs of people on GEDMatch (one-to-one) who are not on
> each others' one-to-many match list but are probably related.
>
> There are essentially two minimum thresholds to set on a 1:1 match:
> segment length, SNPs. I understand segment length, but I really do not know
> to what extent lowering the SNP threshold impacts credibility of a match.
>
>
> Persons A & B =====================
>
> -- I tried a match with segment length (which I will call SL) at 5 cM,
> leaving the SNP threshold at 700. There were no above-threshold matches on
> any chromosome.
>
> -- So I tried a match with SL=5 and SNP=300, and there were matches on
> several chromosomes. The longest segment was 5.4 cM (438 SNPs). The segment
> with the most SNPs was 5.3 cM (1,139 SNPs).
>
> I have these questions from this .....................
>
> 1 - Why did my first run (5 cM, 700 SNPs) NOT find the 5.3 cM (1,139 SNP)
> match, since it is clearly above both thresholds?
>
> 2 - How credible are any of these matches as being due to a common
> ancestor?
>
> 3 - What is the minimal credible SNP threshold?
>
> 4 - What is the minimal credble threshold for both parameters?
>
>
>
> Persons C & D =====================
>
> I tried stepping down the thresholds until I found matching segments.
>
> This was at SL=3 and SNP=300.
>
> The largest segment was 4.1 cM (408 SNPs). The segment with the most SNPs
> was 3.2 cM (529 SNPs).
>
> More questions .....................
>
> 1 - Are these matches credible as being due to a common ancestor?
>
> 2 - Is a higher SNP match more or less credible than a higher segment
> length match?
>
> -------------------------------
> To unsubscribe from the list, please send an email to
> with the word 'unsubscribe' without
> the quotes in the subject and the body of the message
>


This thread: