I'm a computer science student who has started working with DNA. I know the basics but not everything. While working on the ICGC data, I found a weird pattern in the insertions:

In around 60% of the samples in the following setup, I have the INSertion (the "mutation_allele") already in the "ref_context".


  • I have the "ref_context" which contains the reference of where the mutation occured (around 10 nucleotides to the left and 10 to the right)
  • I have the "mutation_allele" which is the mutation that occured (the inserted sequence in the case of insertions).

To analyze this problem in an easier manner, I kept only the insertions of at least length 4.


  • There are 8325 samples where the reference context contains the INSERTION: 60%
  • There are 5383 samples where the reference context DOES NOT contain the INSERTION

A snippet: a snipped of ref_context already containing the INS

I understand that there is a concept of DNA slippage, which can occur during enzymic repair of a repetitive sequence without checking how many repeats should be inserted (and perhaps other reasons). This could be a possible explanation of this. But in 60% of all samples?

I double-checked if I have an error in the ref_context and that somehow the mutation has been already added to it, but no, the ref_context is correct. I queried the position in UCSC and it is indeed a reference.

Another problem is that I have many very long insertions. Is that normal? Out of the 5383 , we have 1511 samples where the len(mutation) > len(ref_context): i.e. 10% of the time. Here is a snippet: very long insertions

  • $\begingroup$ Welcome to SE Biology. Your question is reasonable, so I have gone to the trouble of editing it, but please make a little more effort with English sentence structure, and capitalization of DNA. And it is good if you read to the end of the Tour so you get the badge that shows us you have made the effort to find out how our site works. $\endgroup$
    – David
    Commented Apr 21 at 8:29
  • $\begingroup$ Thank you David, my vocabulary in Biology is quite limited, making it hard for me to express what I mean. I will go through the tour. Thank you for your edit! $\endgroup$ Commented Apr 23 at 9:29

1 Answer 1


Microsatellite instability (which is present in some cancers) is a process mediated by slippage of the kind that you describe.

I am not sure that I exactly understand what is shown in your table (VCF?) or what ref_context is, but mutations from microsatellites are indeed probably a very large contribution to mutational profiles overall, they are merely less famous and often undersampled due to technical constraints, even in healthy people.

I am not seeing a lot of low complexity sequence just eyeballing your variants and contexts, so it is a little funny. Since you only give images I can't do the normal text stuff I would do to check.

A quick way to check this might be to overlay your mutation table with an annotation track of repetitive regions and see whether the mutations are overrepresented there.

Note that many people use repeat-masked genomes when doing mutation calling because it is notoriously difficult to call mutations in repeats, where it can be difficult to distinguish a mutation from a merely mismapped read. Knowing more about your mapping/variant calling process might enlighten us a bit.

  • $\begingroup$ I saved those insertions in a csv excel file, here is a link to it: drive.google.com/file/d/1w3CvcMZ_kR2uXtbdiCiMkLrH6L6dY2bi/… $\endgroup$ Commented Apr 19 at 18:04
  • $\begingroup$ ref_context is just a column in the ICGC data that shows you the neighbors of where the mutation has occured, in the reference genome. The mutation happens in the middle of the ref_context. I didn't undertand the "quick way to check" you said. Where should I get the annotation track of the repetitive regions from ? $\endgroup$ Commented Apr 19 at 18:06
  • $\begingroup$ @WassimJaoui thanks for the link, though it looks like I can't access it. a quick way to check would be to just search short snippets of the context/mutation to see whether they are repeated since it's sometimes hard to see that with eyeballs alone. For the annotation track, UCSC has annotations for basically anything that you could imagine. I don't know which genome build you're working against (GRCh38? T2T?) but they have explicit microsatellite and repeatmasker tracks that you can get at. You can then just run some overlaps with bedtools. $\endgroup$ Commented Apr 19 at 18:33
  • $\begingroup$ Thank you! I will look at that, I'm actually working with an older version: GRCh37 (also known as hg19) I made the link public now: drive.google.com/file/d/1w3CvcMZ_kR2uXtbdiCiMkLrH6L6dY2bi/… Sorry for that error! and thank you very much for the help! $\endgroup$ Commented Apr 19 at 22:06

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .