CNV analysis from exome data (heterozygous vs homozygous)

Question

[younhee.ko]

Hello I have a question to analyze the CNV from exome-data using NGS..

I have family (father, mother, affected patient, non-affected sibling)

I would like to find CNV which is associated with specific disease.

I have used CNVnator and got the list of CNVs.

My questions is how can I filter the CNV which are not associated with specific disease?

This disease is known to be inherited recessive way.. which means that specific CNV we found from patient can also be detected from sibling or father or mother too. In the same way. (even though the ration would be different..?? right?) In other words, since this disease is recessively inherited, patient might have homozygous deletion/duplication on the other hands, other family member might not have such CNV or heterozygous deletion/duplication.. So, in the latter case, disease associated CNV could be detected from all family member.. So, we could not find real CNV associated with a specific disease... How can we handle this problem?

One more!! Can we classify the CNV as homozygous or heterozygous?

Could you please somebody help with this?

Answer 1

[Kizuna]

CNVs which are not associated with your disease are those which do not co-segregate...

So co-segregation filter criteria should be like this :

The disease causing CNV is the one that is HMZ in the index, HTZ in each of the parents. Do not pay attention to the other sibling bcoz his genotype does not give big importance to your analysis, as CNVs are really rare.. so if you find a real one in the index, it should not be found HMZ mutated in the unaffected sib..

if this is a real ar inheritance pattern, your CNV should be HMZ.

However, if your patient is simplex, you might have de novo CNV -> it would be HTZ.

Hope this would help

[Katie D'Aco]

Not true that the sib's genotype is unimportant. Since the sib is unaffected, you want to exclude any CNVs where the he/she is also homozygous.

[Kizuna]

but This is obvious..

However, Logically speaking you can not have a CNV that is HMZ mutated in both sibs.. You are working with a rare disease !!! Every CNV found in two sib is called CNP (copy number polymorphisms) and these variants are not disease causing candidates of mendelian diseases.

sorry to disappoint you :/

The investigator will be lucky if he finds one rare HMZ CNV..

[younhee.ko]

Hm.. my interest disease is not a rare disease.. So, as I mentioned the below comments, I have a list of candidate disease genes (about 200?? ). I would like to identify the specific gene among those large set of gene...... So, as Katie mentioned, I need to filter out with siblig's homozygous CNV..... I think... right?

[younhee.ko]

Hello Kizuna

Thank you for your comments.

So, what you are saying is that ~~

For the disease associated CNVs, (if it is inherited)

1. it should appear as heterozygous deletion/duplication .from one of parents or from both parents
2. it should appear as heterozygous deletion/duplication .from unaffected sibling or not appear from unaffected sibling.
3. it should appear as homozygous deletion/duplication ...

Is it right?

Since I have hundreds of gene list which are known to be associated with this specific Disease already, so I extracted CNVs which are overlapped with any of these disease gene... Now I would like to really find the one!!! which is responsible for this disease using inherited pattern.

BTW, how can I know whether detected CNV is homozygous or heterozygous??

my CNVnator output looks like this.

CNV_type coordinates CNV_size normalized_RD p-val1 p-val2 p-val3 p-val4 q0
duplication     chr1:1-265800   265800  28.9825 0.00202927      0       0.00310948      0       1
deletion        chr1:265801-318000      52200   0.164145        0.0132954       9.92971e+08     23663   2.2852e-148  1

Could you give some comments?

Thanks in advance

Answer 2

[Kizuna]

the co-segregation that I gave is for rare diseases.. if you are working with a complex disease, things are more complex, bcoz you need to perform an association study in family based designs, and here you can not work, on one family.. One family design is for mendelian diseases..

Regarding,

point 1) it should be HTZ in each of the parents...

sorry but I do not use CNVnator to tell you how the genotype should be done.. check the CNVnator article, they should write how to know this issue, if it is reported..

[younhee.ko]

Hello Kizuna,

Thank you for your comments. What do you mean I could not find CNV for the compex disease?

Here is my approach (I think)

1. I have 200 genes that are known to be associated with a specific disease.
2. Using CNV analysis, I found a list of CNV from the patient.
3. find the overlapped CNVs with any of exons of 200 genes.
4. Among the CNV obtained from step3, identify the CNVs that are heterozygous from each of parents.
5. These are list of CNVs we are interested. then, We also report the genes whose exon has overlapped with these CNVs.. It can be one or more.. ( also It will have False positive too.. )but at least they are enough to be candidates.....

Isn't it? Do I miss?? Why we could not do such analysis for the complex disease??

Answer 3

[Kizuna]

I agree with the 4 steps.. However, based on my experience in complex traits, in order to prove a genotype-phenotype correlation you need an association with a alpha 0.05.. and this is not feasible with one family design..

I would advise to search pubmed (e.g: CNV and complex diseases) and find what others did when they wanted to answer the same question.

I hope this would help

[younhee.ko]

Thank you so much!! :) I will try to read some papers related with CNV and complex diseases!!

[Kizuna]

Keep me updated then :)