2
$\begingroup$

Let's say I'm sequencing a donor, and I have 20X coverage. And let's say all the reads covering a particular base has the same allele at the position. Then, is that position considered a "variant" with only 1 allele in that position or is that not even considered a variant?

The reason I'm asking is, I was given some read data, and it's in the form location:variant - location:variant.

But there was often the case where for all the variants at a particular location, there was only one allele. But this data is labelled to only have variants. Hence, how is a particular location deemed a "variant."

For example, lets say I'm told that the following is "variant reads given in the form read# location:variant - location:variant"

read 1 6523432: A - 6710432 T
read 2 6523432: C - 6710432 G
read 3 6523432: A - 6710432 T
read 4 6523432: C - 6710432 G

Then, why are these two locations considered variants? It seems like it's just a heterozygous locations. Why were these locations considered to carry variants? Similarly,

read 1 6523432: A - 6710432 T
read 2 6523432: A - 6710432 T
read 3 6523432: A - 6710432 T
read 4 6523432: A - 6710432 T

In those reads, it just seems like it's homozygous. Why is it being called a variant? How was it classified as such?

Put another way, if I'm given 100000 reads, and I said I only want to look at informative reads, and so I'm given a subset with only 100 reads, and those 100 reads look something like the above, then how were these reads selected?

$\endgroup$
  • $\begingroup$ Please edit your question and include an example. I am afraid it is very hard to understand what you mean. I suspect you are using the terms variant and allele in a non-standard way, so an example illustrating what you mean exactly would be really helpful. $\endgroup$ – terdon Apr 17 '19 at 16:35
  • $\begingroup$ @terdon I've editted my question. $\endgroup$ – Jonathan Apr 17 '19 at 17:39
  • $\begingroup$ Whether you call something a variant or not depends on what you are using as a reference. You need the genomic position, not the position with respect to the read and you need a reference genome to treat as a "truth set". You cannot know whether something is a variant or not just by looking at the reads without a reference genome. What kind of data are you analyzing? Have you aligned your reads to a genome? $\endgroup$ – terdon Apr 17 '19 at 18:11
  • $\begingroup$ I see. These position, I know, are positions on the genome. Hence, I'm guessing these reads have been mapped to the reference, and then being called variant. Is that correct? $\endgroup$ – Jonathan Apr 17 '19 at 18:28
  • $\begingroup$ So, in the 2nd read cloud, if the reference was heterozygous, but the donor reads are telling me it's homozygous then it's called a variant. Am I understanding that correctly? $\endgroup$ – Jonathan Apr 17 '19 at 18:28
3
$\begingroup$
  1. Set 1

    read 1 6523432: A - 6710432 T
    read 2 6523432: C - 6710432 G
    read 3 6523432: A - 6710432 T
    read 4 6523432: C - 6710432 G
    

    I understand this as showing the location of 2 variants, supported by 2 reads each. Now, if the reference has an A at position 6523432, your sample could be heterozygous for an A->C variant at position 6523432. If the reference has a C at that position, then your sample could be heterozygous for C->A. If the reference has something else again, for example G, then your sample could be heterozygous for G->A/C. Similarly for location 6710432.

  2. Set 2

    read 1 6523432: A - 6710432 T
    read 2 6523432: A - 6710432 T
    read 3 6523432: A - 6710432 T
    read 4 6523432: A - 6710432 T
    

    In this case, if the reference has an A at position 6523432, the reads do not support a variant. However, if the reference has anything else, then the reads would support a homozygous variant e.g. G->A. And the same for position 6710432.

Now, in all of these cases, you cannot know whether a variant is real without taking into account the number of reads that support it, the quality of the alignment, and the quality of the reads. That's why we do this using dedicated variant callers, software that is designed to take all these factors into account and decide whether to call a variant at that position.

But in all cases, the variant must be defined with respect to a reference genome. You cannot call a variant just by looking at reads without any knowledge of a reference. A "variant" just means "not what is found in the reference". It is largely a matter of random chance whether something is a "variant" or "normal". If the individual(s) whose genome was taken as a reference has the same sequence, then you call it normal. If not, then you call it a variant. There are many cases where the standard hg19 or hg38 genomes have a "normal" residue which is actually far less common than the known "variant". There's nothing special about the reference genome, after all, it's just the one we chose to call a reference.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Perfect, this exactly the clarification I needed, thank you! $\endgroup$ – Jonathan Apr 17 '19 at 20:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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