4
$\begingroup$

I'm trying to play around with my family's raw genetic data from Ancestry.com so that I can do some genetic analyses, and I noticed in the AncestryDNA raw data files that some regions of the genome have more than one variant associated with them and I would like to know why this is. The end goal is get this into a regularly formatted VCF, but I feel that these double variants will probably mess things up. So for example one might see something similar to this in the tab-delimited file:

rs6010164      22    50515270    T    T
rs267607236    22    50515270    I    I

(Header line: rsid chromosome position allele1 allele2)

I understand that when rsid numbers were created, sometimes multiple rsid numbers were accidentally assigned to the same region of the genome, but I do not understand how two separate variants can be assigned to one single region in the genome. I could be wrong but as I understand it, I should only see one line here (either a "T T" or an "I I") because genotypes are reported in comparison to the human reference genome and so each position should uniquely refer to a single nucleotide in the said human reference genome.

But in this example case one line says there is a "T T" genotype at this position, and the other line says there is an insertion(see "How do I read my raw DNA data?" here). One possible explanation I thought of is if this would just be saying that a T was inserted, but if this is the case, then as I understand it the reference allele present in the individual should still reported, just in conjunction with the allele at the index. (So assuming an extra "A" was inserted after the T, which could be the reference or alternative allele, something like this should be recorded: rs6010164 22 50515270 TA TA).

Can anyone help me understand why there are two separate variants associated with chromosomal position? Thanks!

$\endgroup$
5
  • 2
    $\begingroup$ Are you sure that if an A was inserted, you would get GA GA? I have seen G GA for an A insertion. Looking back at your header, it does not seem to respect the VCF format. The order of the first three columns is scrambled. It should be chromosome, position, id. Then, there should be the reference sequence at that position, followed by the alternative sequence. Then, in the sample column, you should get the genotype in function of the reference and alternative alleles. $\endgroup$ Mar 13, 2018 at 1:34
  • $\begingroup$ Hi @charlesdarwin yes you are right its not a regular vcf, I should probably just refer to it as a tab-delimited file. No I am not sure that I would get that, I am not entirely sure what format indels are supposed to be in, but I do feel like its super weird for them to have more than one line associated with a single position, and eventually I need to change this into a VCF file and I feel like that would probably mess things up. $\endgroup$
    – Sarah
    Mar 13, 2018 at 3:39
  • 3
    $\begingroup$ Did they perform exome sequencing or use a SNP chip? This is likely to occur for the latter, since they're then reporting every probe. $\endgroup$
    – Devon Ryan
    Mar 13, 2018 at 8:09
  • $\begingroup$ @DevonRyan I believe they used a SNP chip. $\endgroup$
    – Sarah
    Mar 13, 2018 at 19:01
  • $\begingroup$ The VCF format does allow multiple variants at the same chromosome & position - in fact the official dbSNP VCFs have exactly that. $\endgroup$ Nov 12, 2020 at 9:41

4 Answers 4

2
$\begingroup$

First of all, take these data with a very large pinch of salt. This sort of targeted analysis is not designed to produce high quality genetic data but, to give an idea of a sample's ancestry. Given a file like the one you describe with no quality information, you have no way of knowing how reliable any of the variants called actually is. From personal experience with another, similar company, I expect that they aren't reliable at all.

I have tested four individuals using a commercial tool like the one you describe and while I don't doubt the validity of their main results (those displayed on their website), their raw data showed combinations of variants that were extremely unlikely to be real given that the individuals were still breathing.

This isn't to say that such companies don't do their job well, only that their job is limited to what they actually display. When you look into their raw data, you're on your own. Presumably, they apply various filters before displaying the main results, but these are most likely not applied to the raw data. So if you want to do your own analyses, this is unlikely to be a reliable dataset. Also, I would strongly urge you to seek the help of an expert when interpreting any data you do have.


That said, the rsIDs are simply identifiers of a specific genetic variation, not of a location. Any genomic position can have many (essentially infinite) possible variations. For example, a C residue at position 100 of chromosome 1 could become a G, a T or an A. Each of these will have its own, separate rsID. Now, if you also consider insertions and deletions, you can see why you can have essentially limitless possible variations at that specific position. So no, having more than one rsID describing the same position is not strange at all.

The way these tests work is they look at a list of possible variants, each of which has its own rsID, and they report what they find for that rsID in the sample. Then, they will report what they find at the rsID's position for each allele. This is explained in the FAQ article you linked to:

The possible observations are A for adenine, C for cytosine, G for guanine, T for thymine, I for insertion’ and ‘D for deletion, or 0 for missing data. Column one provides the identifier (including the #rsID where possible). Columns two and three contain the chromosome and basepair position of the variant using human reference 37.1 coordinates. Columns four and five contain the two alleles observed at this variant (genotype)

Let's have a look at the two rsIds you mention.

  1. rs6010164: This is a single nucleotide variant (SNV) which describes a C residue where the reference genome has a T.

  2. rs267607236: this is a deletion at position 50076841, so a very different variant to the above. In this one, the sequence TGAG is deleted.

Your results are:

rs6010164      22    50515270    T    T
rs267607236    22    50515270    I    I

So, your sample doesn't have rs6010164 since it has the same sequence at that position as the reference. However, it does have the insertion as indicated by the two Is.

The Is and Ds of the insertions and deletions (indels) are going to be the easy bit here. The only way you can parse this sort of file and convert it to VCF is to get the list of target SNV rsIDs, check what they describe as the reference and as the variant and then, for every one, check which of the two is shown in your file.

If you're not an expert in this field, you will probably be better off hiring someone to do this for you. There are actually companies that offer this sort of service and can take these tab-delimited files as input. I even work for one!

$\endgroup$
2
$\begingroup$

It's pretty common for SNPs and indels that map to the same position in the genome to be assigned different rsIDs, particularly when the indel spans multiple base pairs in the reference. You can get more info about these particular variants here and here. Or you can see how both of these variants relate to the reference sequence here.

Without info about the reference allele it's pretty confusing (particularly because they appear to be reporting the opposite strand from the database (G vs. C/T), but I would interpret this to mean that both alleles match the reference for rs267607236 (they have the insertion), but do not match the reference for rs6010164 (G-C pair rather than T-A). This is also the most common genotype at this position, being found in almost 80% of people in most populations.

$\endgroup$
1
  • $\begingroup$ Thank you @heathobrien! If I want to convert my data to a VCF don't I need to one use only one of the rsid's since they are at the same position? Is there any standardized way to pick which one to use? And sorry, my bad, they did report from the forward strand. I had changed the nucleotides since its actually not my own DNA data I'm posting here, but I should have checked to make sure I was changing it to a genotype that made sense. $\endgroup$
    – Sarah
    Mar 13, 2018 at 19:05
2
$\begingroup$

You need to understand a few key points for your example given:

(a) Ancestry (and 23andMe) will use the letters I and D to indicate it is an InDel/DELINS. You have to read the rsID definition to determine if the variant is an Insert or Delete from the reference. And therefore which letter is used to mean positive / variant and which for negative / reference. They simply use the other letter to indicate negative / reference.

rs267607236 is defined as a deletion. So the I's mean the reference was detected in this case. Therefore the Ancestry file indicates no deletion of the 4 base-pairs at coordinates 50515271-4.

(b) Ancestry and dbSNP VCF annotation files uses a "left-aligned" coordinate for ALL InDel's and generally designate the coordinate BEFORE the variant.

This makes sense for Inserts. It gives the coordinate of the last base-pair in the reference just before the insert starts. That coordinate plus 1 is the next base-pair after the insert in the reference.

This same method is used for deletions. But because the deleted values have coordinates in the reference, this makes deletions appear "one off". You could use the starting coordinate of the deletion when defining the variant (e.g. [TGAG/-]). Otherwise, you need the coordinate before if defining the deletion as [CTGAG/C].

rs267607236 is listed in the OP's information as coordinate 50515270. Because it is a deletion, the deleted sequence actually starts with the base-pair at coordinate 50515270+1 or 50515271.

(c) As it ends up, for rs267607236 in question here, the dbSNP VCF and the dbSNP online website entry are different (both latest release 156 in this case). This could happen but still be consistent because there is a T flanking the 4 base-pair deletion. The starting reference sequence ...C TGAGT A... becomes CTA after the deletion. But which T remains is not really known. The VCF file defines the deletion by retaining the trailing T. The online website retains the leading T.

And in fact, it appears the website entry starting coordinate is incorrect. Given the way they defined this variant within the flanking sequence, the online website should use the coordinate 22:50515270 (like in the Ancestry file) and not 22:50515271 as it does now. This may have existed with that coordinate in an earlier release used by Illumina / Ancestry when defining the pipeline. But why the VCF and online website are different now is confusing.

There are numerous other nuances, conventions and actual errors that have to be understand to properly convert a microarray raw data file to a VCF.

Here are all the entries that we referenced / consulted brought together for your perusal:

rs6010164      22    50515270    T    T
rs267607236    22    50515270    I    I

GRCh37 Reference sequence around rs267607236:
TTAC TGAGT AAGA  (70-74)

dbSNP Release 156 GRCh37 VCF:
NC_000022.10    50515270        rs6010164       T       C
NC_000022.10    50515269        rs267607236     CTGAG   C       # Note: leading T deleted

dbSNP web page definition:
Alleles:T>C             for rs6010164)
TTAC [T/C] GAGTAAGAT
22:50515270 (GRCh37)
NC_000022.10:50515270:T:C

Alleles:TGAGT>T         for rs267607236
TTACT [GAGT/-] AAGAT                                            # Note: trailing T deleted
22:50515271 (GRCh37)                # for consistency with flanking sequence, this should be ..270
NC_000022.10:50515271:TGAGT:T

It is not clear to me what Ancestry does when the same rsID has an insert and a deletion at the same position. For example, the line from dbSNP VCF for rs768019142 is:
NC_000001.11 10055 rs768019142 TA T,TAA
which says there are two variants in the rsID. (This is common, especially for SNPs with multiple variants possible). A deletion at 10056 OR a single letter insert after 10056. To represent this, instead of a single letter for the reference, they have given two base-pairs as the reference sequence. They probably should have defined two rsIDs at two adjacent locations. UCSC's versions of the VCF files are filtered and cleanup to simplify items like this.

$\endgroup$
1
$\begingroup$

For different variation type at the same location, each variation type is being assigned its own rsID by the dbSNP maintainers.

In your example, one chromosome in the pair has a single-nucleotide variation (T allele), while the other chromosome has an insertion variation.

Quote:

dbSNP only merges rs numbers that have an identical set of mappings to a genome and the same allele type (e.g. both must be the same variation type and share one allele in common). We therefore would not merge a SNP and an indel (insertion/deletion) into a single rs number (different variation classes) since they represent two different types of mutational "events".

In your case, rs6010164 is of the Single Nucleotide Variation (SNV) variation type, while rs267607236 is of the Small deletions or insertions (DELINS) variation type.

Quote:

Despite its name, RefSNP is assigned to all variation types listed below with precise locations for both common and rare variations, including mutations. Most are typically small variations (<= 50bp).

  • Single nucleotide variation (SNV)
  • Short multi-nucleotide changes (MNV)
  • Small deletions or insertions
  • Small STR repeats
  • retrotransposable element insertions
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.