Genetic studies of ME/CFS and other diseases including GWAS - discussion thread

[Jonathan Edwards]

I think known, but yet undiagnosed genetic disorders, may be able to be identified and this would be useful for some patients.

Are they planning to do precision work on people in the DecodeME sample?

Eg. Based on severity or co-morbid groupings?

As I understand it, and I may be wrong, the GWAS looks for links to polymorphic sites along the chromosomes that should show correlations with gene polymorphisms in the local area if they confer disease risk. So it can pull out effects of genes not so far known to be associated with disease. It does not require prior knowledge of all the polymorphic forms of every gene that might be relevant.

Identifying the gene actually linked is then another task but a relatively easy one once the polymorphisms have been documented.

Again I may be wrong but my understanding is that the constraining factor in getting any useful result is the statistical power of any association found with anything but a very large number of subjects. Subsetting at this stage almost defeats the object of the primary search.

However, if a risk signal is found looking at subsets then becomes practical. It also becomes much easier for other databanks to replicate and look at subsets.

 

[Hoopoe]

I think known, but yet undiagnosed genetic disorders, may be able to be identified and this would be useful for some patients.

Are they planning to do precision work on people in the DecodeME sample?

I vaguely remember that this is not planned and believe this is because the technology used is different from that required for diagnosing a genetic disorder. Maybe also because it would be a lot of extra work.

 

[Amw66]

As I understand it, and I may be wrong, the GWAS looks for links to polymorphic sites along the chromosomes that should show correlations with gene polymorphisms in the local area if they confer disease risk. So it can pull out effects of genes not so far known to be associated with disease. It does not require prior knowledge of all the polymorphic forms of every gene that might be relevant.

Identifying the gene actually linked is then another task but a relatively easy one once the polymorphisms have been documented.

Again I may be wrong but my understanding is that the constraining factor in getting any useful result is the statistical power of any association found with anything but a very large number of subjects. Subsetting at this stage almost defeats the object of the primary search.

However, if a risk signal is found looking at subsets then becomes practical. It also becomes much easier for other databanks to replicate and look at subsets.

I am not up on genetics but I have a recollection of Chris Ponting confirming the extent of the study re number of genes on a video. It's not full GWAS probably due to cost constraints : perhaps @Andy can advise ?

 

[Andy]

I am not up on genetics but I have a recollection of Chris Ponting confirming the extent of the study re number of genes on a video. It's not full GWAS probably due to cost constraints : perhaps @Andy can advise ?

I believe you are thinking of the difference between a GWAS, which we are doing, and whole genome sequencing, which we are retaining part of each donated sample in the hope that we will be able to do in the future. Any GWAS only looks at selected areas of the genome, "at approximately one million locations where people are most likely to differ", https://www.decodeme.org.uk/the-science/, whereas whole genome sequencing looks at all 3 billion locations in the human genome. Whole genome sequencing costs a lot more than a GWAS, as you can probably imagine.

 

[Amw66]

I believe you are thinking of the difference between a GWAS, which we are doing, and whole genome sequencing, which we are retaining part of each donated sample in the hope that we will be able to do in the future. Any GWAS only looks at selected areas of the genome, "at approximately one million locations where people are most likely to differ", https://www.decodeme.org.uk/the-science/, whereas whole genome sequencing looks at all 3 billion locations in the human genome. Whole genome sequencing costs a lot more than a GWAS, as you can probably imagine.

Thanks @Andy . Exactly this.

 

[Jonathan Edwards]

Any GWAS only looks at selected areas of the genome, "at approximately one million locations where people are most likely to differ",

Is that quite right, Andy? Or at least perhaps misleading. My understanding is that GWAS is called genome wide because the mutation sites looked for cover the whole genome. They do not deal with every mutation in every gene but the technique relies on polymorphisms carrying association with variations in putative pathogenic genes through linkage disequilibrium. That means that if there is a risk associated with a gene X mutation it should be picked up by a link to one of the GWAS targets (?SNPs) nearby. Clearly, looking all individual polymorphisms in individual genes would be more precise but the problem of correcting for multiple outcomes would be even bigger so would require even huger cohorts?

The issue is whether or not a GWAS with a million targets is likely to miss an important genetic link. My understanding was that the risk of that is fairly low, but I may be wrong. After all we only have 25,000 genes as far as I know.

 

[Andy]

[Struggling even more so with brain fog at the moment, but I'll do my best to answer.]

Well, obviously I am not the expert in GWAS, so I am relying on other people who understand the detail better than I to explain things.

@Simon M wrote a blog about GWAS, thread on it here Researchers propose deep trawl of DNA to help uncover the causes of ME/CFS (Simon McG blog), where he explains

"In fact, GWAS typically screens only around 10% of all common SNPs because its immediate neighbours, the ones on the same block, will behave in a virtually identical way. Including neighbouring SNPs in a GWAS will add very little information.

In practice, the SNPs in a GWAS serve as genetic markers, each tagging a block of DNA."

which I think covers the core of your question. As to whether there is a chance that DecodeME, like any other GWAS conducted, will miss an important genetic link; it is not zero but neither is it considered to be high.

I've asked the Genetics Delivery Team of DecodeME for a suggested article or paper that covers the rationale of screening a selection of SNPs as opposed to all SNPs, I'll report back on that.

 

[Jonathan Edwards]

"In fact, GWAS typically screens only around 10% of all common SNPs because its immediate neighbours, the ones on the same block, will behave in a virtually identical way. Including neighbouring SNPs in a GWAS will add very little information.

In practice, the SNPs in a GWAS serve as genetic markers, each tagging a block of DNA."

Yes, this is basically what I was suggesting. The GWAS does not look for every known polymorphism at the individual base level (SNP) but it is designed to cover every part of the genome so that linkages should be picked up because the SNPs that the GWAS does cover are quite closely spaced in comparison to whole genes.

It may not be quite the case but if we have 25,000 genes and the GWAS looks at a million SNPs then each gene (with associated promoter, introns and junc sequences) should be covered by 40 hits? Typically, results from GWAS show a linkage attributed to possible variants of more than one gene because segments of several genes tend to be quite tightly linked (i.e. in disequilibrium).

So for instance, for the haemoglobin gene the GWAS may to cover the site of sickle haemoglobin itself but will have about forty tags within the gene domain and relevant polymorphisms are highly likely to be linked to some or all of these.