Target deletions with adaptive sampling

Question

[njornet]

I want to perform adaptive sampling to detect a panel of diseases. Some of them are caused by deletions such as Prader-Willi syndrome and Angelman syndrome, both are cause by the deletion of 15q11.2-q13. How should I design the bed file to target this region? Do I include the whole region?

Edit: For more context, we want to detect these diseases in fetal cells from a mother's blood sample, so we will have a small amount of fetal DNA. At the current state we are performing whole genome amplification, so we won't be able to detect the diseases if they are cause by imprinting defects. From my research, PW is cause by a deletion of the 15q11.2-q13 region and Angelman by 15q11.2-q13 deletion in about 70% of cases.

So in general, what I'm asking for are tips on how to detect deletions using AS.

Answer 1

[LauferVA]

I want to perform adaptive sampling to detect a panel of diseases.

this is the most problematic statment of your post. how many loci. what are they? how big is each? the feasibility of your study may well depend on what you mean by this nebulous statement...

Some of them are caused by deletions such as Prader-Willi syndrome and Angelman syndrome, both are cause by the deletion of 15q11.2-q13.

.... this is not entirely accurate.

while deletion followed by restoration of normal copy number through uniparental disomy or another mechanism is certainly one possible pathogenetic mechanism of PW or Angelman, it isnt really accurate to say that they are caused by deletions without further qualification... they may be also caused by issues with imprinting without evoking other mechanisms, or may be simple deletions i beleive... in a minority of cases....

but the main reason why nanopore could make sense is that these are disorders of imprinting. the use of nanopore is not needed to detect a mere deletion - this can be accomplished with far more ease through other technologies... rather, the utility of nanopore in this context flows from its ability to detect epigenetic modification primarily, and its ability to capture SVs secondarily.

How should I design the bed file to target this region?

This question is difficult to answer without greater clarification as to your goals. generally speaking, for AS and PW much of the epigenetic modification is concentrated into so-called "imprinting control regions" (ICRs) or differentially methylated regions (DMRs). you would probably want to target these regions first and foremost... but in practice the amount and patterning and methylation varies by tissue type, environmental factors, genetic background and everyone's favorite, "stochastic events" ... so its pretty difficult to give you meaningful advice on what you would want to do without further clarification.

Do I include the whole region?

you need to make a detailed study of this region, including genetic variation in the region in your population. the ICRs are thought to be separated by a few hundred kb, and the whole region is - i if understand correctly, 3Mb or so... so if that is all you need, its honestly a pretty good size for a AS approach. other labs (e.g. Mayo laboratories) has reported pretty nice enrichment of a 3Mb region (up to 150x) at AMP last year. generally 3Mb is 0.1% of 3Gb which is on the low end of what nanopore recommends as an "optimal" target region size for AS.

so overall the approach sounds like like it could be feasible, and including the whole region might be easier that trying to just specify the regions containing the ICRs only - that seems likely to result in loss of relevant reads that begin outside the region, but extend into the regions you might care about. in general, yes is probably the answer to this question but again. you need to do a lot of legwork to know this is true for you....

---

the bottom line here is you are going to need to make a serious study of this locus and any other loci on your panel before doing this, or find published reports of others doing the exact same things to use as an initial template.

at any rate, i hope that my post gives you enough vocabulary and what have you to get you started...

[njornet]

Thanks for the answer. You are right I didn't give enough context. We want to perform screening tests on fetal cells found in the mother's blood, so we will probably need to do WGA. I'm aware of the imprinting mechanism causing these diseases, but for now, until we are able to extract enough fetal cells, we won't be able to detect that, only deletion based mutations (or sequence variations for other diseases). We will try to get to that point, but from what I've seen, the most common mechanism behind these two diseases is deletion.

We want to study a couple more deletion based disease (as far as I'm aware only these two are also cause by imprinting issues) and some monogenic diseases, my rough estimations are that we will be targeting around 2% of the genome including these whole regions of the deletion caused diseases.

Again, thank you for your detailed answer, it is really helpful.

[LauferVA]

at the outset i would define the entire 3mb region. AS gives diminishing returns if you try to define things too tightly. remember, you need to pad every area of interest anyway.

i am not sure if what you are saying about base modification is true. the signal to noise for base modification isnt that much worse than base-to-base calling.

doing adaptive sampling to target a deletion may not be the best idea. you would be rejecting essentially all your reads in the event of a positive case.

2% is good for AS, but only if you have relatively few target regions. AS doesnt work as well if you have a ton of tiny little target areas due to the padding issue

[njornet]

What do you mean by this "the signal to noise for base modification isnt that much worse than base-to-base calling."?

[LauferVA]

no problem. please accept the answer if you believe it is answered. this helps future folks who want to refer to the information

Answer 2

[Brian Bushnell]

You are posing a question that assumes a lot of prior knowledge, because it is very domain-specific. If you want a useful answer, I suggest you rephrase the question in a way that does not require prior work in your lab to answer it. For example, mentioning the target organism, the chromosome, the region of interest (in bp), and what software you are trying to use, would all be useful.