In Variant Detection, RNA-sequencing, and ChIP-sequencing, we showed how we matching, or aligning, short NGS reads to the genome, and then make inference using the reads which aggregate or "pile up" at a given region: either to find a SNP, to quantify the amount of mRNA, or the quantify the amount of a protein which is bound to DNA. Define a "true positive" as detecting a SNP, detecting mRNA expression, or detecting a protein which is bound to DNA. Which of the following (you can pick more than one) could possibly lead to a false negative, e.g. we do not detect a true SNP, mRNA expression, or a bound protein at a location/region?

which one is the correct answer? why? (we can pick more than one)

1.The region we care about is not in the reference genome

2.The region we care about occurs 1000s of times in the genome, and we are ignoring reads which align to so many places

3.For the region we care about, the organism's genome is so different to the reference genome that the aligning program can't find a match

4.There are never false positives

  • 2
    $\begingroup$ Homework questions are received well when the asker shows an attempt to answer and explains their confusion. Which answer(s) do you think is (are) correct, and why? $\endgroup$
    – acvill
    Oct 12 '20 at 13:44
  • $\begingroup$ I think that the second and third ones are correct. because when the region occurs 1000s of times, the amount of aligning reads to each part results in decreasing numbers of possible aligning reads to only one specific region. and in addition, when the organism's genome is so different from the reference genome it could lead to a false conclusion of the alignments of the reads. I am not sure what is the difference between the first and third options!!? could you please help me with that? if my answers are incorrect ? $\endgroup$ Oct 12 '20 at 14:31

An example of 1) might be if you were looking for a mutation on the Y chromosome, but your reference sequence had only autosomes.

Read #2 a bit more carefully. If there is a region repeated 1000 times in the genome, a lot of aligners are just throwing those reads away, on the grounds that you can't learn anything useful from them, and they will uselessly bloat the output files. Diluting the coverage by splitting the read coverage will happen with some aligners if there are just a few places for the read to go, but not 1000.

Edit. Let me try and make a better example.

You have a sample of an antibiotic resistant strain of some nasty bacteria. This bacteria has some genes located on a plasmid. Some other lab sequenced this bacteria and its plasmid; that's your reference.

You align your reads to your reference strain of bacteria, but oops, you accidently omitted the plasmid. So your variant caller spotted no differences located in the plasmid. If you were hasty and didn't check to see read coverage on the plasmid, you might assume that your sample's plasmid matched the reference, but you accidently didn't check that.

Next you get another antibiotic resistant sample. You align it to your reference, and see no variants. Okay, then, that's that...except that the reason you see no variants is because the people who gave you the sample misclassified it; it's really a different sub-species, and some regions of your reference are too different from your sample for the reads to align.

In general, you just have to remember that if your variant caller says there's no variant somewhere...that could also means there are no reads there. In real life, people will do some QC on the read alignments stats to try and catch that.

  • $\begingroup$ so in your opinion, the first and the second one are the correct answers? $\endgroup$ Oct 12 '20 at 19:20
  • $\begingroup$ would you please explain about the third one? I cannot recognize the difference between this and the first one. $\endgroup$ Oct 12 '20 at 19:22

The first three choices can and do occur in genomics. For examples of each: genomes of cancerous cells involve translocations, which can create regions of the genome which do not occur in the reference genome. There are regions which occur 1000s of times in a genome, called transposons. Many software pipelines discard reads which map to many locations, and then true biology (e.g. a bound protein) which can occur at those regions is ignored. Also, in sequencing RNA or DNA from cancer cells, there can be so many mutations that the aligning programs have difficulty finding the match in the reference genome. Or, this can happen if we align reads from one species to the genome of another species (which does not yet have a reference genome constructed).


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