Preprint Initial findings from the DecodeME genome-wide association study of myalgic encephalomyelitis/chronic fatigue syndrome, 2025, DecodeMe Collaboration

A select number of answers from the FAQs that might address recent discussion points.

Q. Do these results mean everyone with ME/CFS has one or all of these genetic differences?

A. No, these findings are based on large-scale patterns across the thousands of people with ME/CFS that we studied. They show that, on average, people with ME/CFS are more likely to have these differences than people without ME/CFS. These signals are about what we can see across populations, not individuals.


Q. Are these findings unique to ME or have they been found in other illnesses?

A. The signals we have found are different from those found in other illnesses to date, except for the one on chromosome 17 that was previously found in people experiencing chronic pain.


Q. Can these results lead to treatments for ME/CFS being developed?

A. DecodeME’s purpose was to gain an understanding of the biological roots of ME/CFS and how genetics could contribute to developing the condition. By identifying genes linked to the immune system and the nervous system, the study opens the door for scientists to explore what’s going wrong in ME/CFS at a molecular level. This deeper understanding could, in time, lead to treatments that target the root causes of the illness.


Q. What happens next with the study?

A. Although the initial results of DecodeME are now available, we are not finished. We’ll continue to analyse the genetic data, and we will update our scientific paper as needed before it is peer reviewed and published. We also have a detailed and valuable dataset from the second questionnaire that focused on symptoms, quality of life, treatments and therapies. We will be analysing and reporting on this in due course.

Our work doesn’t end there. Approved researchers will be able to use the data of those who consented through our data access process, helping to catalyse new studies and discoveries.


Q. What happens next for ME/CFS research?

A. The DecodeME findings are a springboard for future ME/CFS research. Now that we’ve identified genetic signals linked to ME/CFS, researchers can start investigating exactly what’s happening in the body – for example, by specialist scientists delving deeper into each genetic region found. These discoveries open new avenues for research, potential drug development or repurposing, and targeted treatments. DecodeME’s findings have laid the groundwork for a new era of ME/CFS research, and our rich dataset is available for data access applications and is already driving new discoveries forward.
 
Last edited:
In an interview with David Tuller, Ponting also said something interesting (starting at minute 21:23):
Ponting: What we are doing is asking the question whether the associations we are seeing are equivalent to what others have seen for other diseases. And thus far the ME ones seem to be specific to ME and not to any other traits or disease aside from this one on chronic pain.

Tuller: And that means what?

Ponting: It means that the ME genetic signals are not equivalent to any… arthritis, Parkinson’s, Alzheimer’s, depression, anxiety, none of those and more.
 
In an interview with David Tuller, Ponting also said something interesting (starting at minute 21:23):


Could you, based on those analyses, say to HCPs that they have to stop thinking about ME/CFS in terms of the diseases they know about, because the genetic signals indicate that it doesn’t work like any other known diseases? Or am I over-interpreting what Ponting said?
 
In the data analysis plan there was the suggestion of combining the DecodeME and UKB ME/CFS cases to get more power.

Was this done or was the idea that the UKB ME/CFS cases are too unreliable?
It hasn't been done yet, though something Chris said on the video makes me think they are considering it.

A big problem is that, while they can pick up people with a recorded diagnosis, we know many of those are wrong. And there's no questionnaire data to qualify.

The paper refers to the three "pen" questions – but these are about fatigue and nothing like the PEM question asked by. DecodeME of all participants (we also asked a lot of other questions to qualify people currently having ME/CFS )

And I think the numbers are only about 1000 for recorded diagnosis plus the three questions they used.
 
Could you, based on those analyses, say to HCPs that they have to stop thinking about ME/CFS in terms of the diseases they know about, because the genetic signals indicate that it doesn’t work like any other known diseases? Or am I over-interpreting what Ponting said?
Somewhat! It's basically saying that if someone hypothesized ME/CFS was just depression by another name, you would expect some relevant overlap which we don't see here. The lack of overlapping hits doesn't completely exclude that possibility (because genetics results really don't have that power), and there may still be overlap in relevant pathways that were simply missed in the study for a variety of reasons. But it means that there's no positive support in favor of "ME/CFS == depression" from the genetics.
 
In an interview with David Tuller, Ponting also said something interesting (starting at minute 21:23):
I believe the inability to find an existing drug that works for ME/CFS also suggests that ME/CFS is quite different from other diseases, or at least the ones that look similar enough to make someone believe a treatment used for another disease would be worth trying in ME/CFS.

The alternative explanation was that ME/CFS didn't really exist and a cohort of ME/CFS was just a collection of misdiagnosed cases and other diseases that medicine can't yet explain. DecodeME has made that a lot less credible.
 
Last edited:
Could you, based on those analyses, say to HCPs that they have to stop thinking about ME/CFS in terms of the diseases they know about, because the genetic signals indicate that it doesn’t work like any other known diseases? Or am I over-interpreting what Ponting said?

I think this is probably an overinterpretation. Two diseases could operate in almost exactly the same way, with a particular sort of disruption in a pathway in a particular type of process - endocrine or neural maybe - but making use of different specific gene products. Haemophilia A and haemophilia B are essentially exactly the same type of disease and it is fine to use one as an analogy of the other but there is no overlap in their risk genes - which are factor VIII and factor IX.

The comment was probably made specifically in the context of rumours that genes would overlap with neurotic illnesses. That they don't is significant but the same argument applies.
 
Edit: For the variant that isn't in GeneAtlas, I see it on this other website that just shows the allele frequency in the BioBank: https://afb.ukbiobank.ac.uk/variant/chr13-53194927-GT-G Maybe it just wasn't compared to CFS in the BioBank for some reason. But the data is still there for a few hundred thousand people.
Thanks for the link. That database was generated from the WGS database according to the about page. LINK

The Supplementary Methods doc states this which might be a clue.
In addition to the genotyping quality-control that UKB performed (14), we removed variants that showed: (i) discordance between our own and UKB’s genotype calls in a batch of 4,700 UKB samples, or (ii) a large departure of allele frequency in European-ancestry UKB samples
from that reported in GnomAD (v2.1.1) for non-Finnish Europeans.

We declared discordance when less than 95% of calls for a variant were concordant. We tested significant allele frequency differences assuming a binomial distribution of allele count, discarding any variant more than 6 standard deviations from the expected mean.
So the control data QA sweep used GnomAD (v2.1.1) which is an hg19 reference. But the DecodeME paper references hg38 locations?
 
In the data analysis plan there was the suggestion of combining the DecodeME and UKB ME/CFS cases to get more power.

Was this done or was the idea that the UKB ME/CFS cases are too unreliable?
I saw no mention of it when I scanned the Supplementary Methods section. They did go through the control data and remove any hint of people with ME/CFS in addition to those labelled as CFS. LINK
 
In an interview with David Tuller, Ponting also said something interesting (starting at minute 21:23):


Does that finally answer the long-standing debate about whether it's the same as fibromyalgia then?

Or did a fibro GWAS study also report finding the CA pain gene?

Eta sorry but I'm too severe to watch the video at the moment. @Andy Apologies I did actually try to look at the FAQs but on my mobile only a couple results ones show up and I couldn't see the answer.
 
Last edited by a moderator:
Does anyone know what is the genetic vs environmental influence they’ve estimated for ME given these new GWAS results? I swear I read somewhere that it’s estimated to be only a ~10% genetic contribution? Can’t find the source again

And if it’s predominantly environment that influences ME wouldn’t a large scale EWAS study make more sense than further genetic studies?
 
Something that hasn't been discussed much is the low heritability estimate
We estimated ME/CFS SNP-based heritability from GWAS-1, based on the LDSC method and reported on a liability scale. It was modest but significantly different from zero, with ℎ^2 = 0.095 (SD = 0.006).
This was estimated using linkage disequilibrium score regression ( LDSC). If I understand correctly the method is relatively simple: it does a regression analysis of linkage disequilibrium (how much SNPs correlate with each other) and test-score statistics for GWAS (how much different the variant was in ME/CFS compared to controls). The heritability estimate is then the slope of this regression line. If there is high heritability, then SNPs with high LD would have higher test statistics and lower p-values.

It seems that there are some problems with this method
The accuracy of LD Score regression as an estimator of confounding and genetic correlations in genome-wide association studies - PMC

And it also doesn't take the sex chromosomes and rare variants into account.

But if we take it as a rough estimate, it still seems quite lower than what many expected?
 
Does that finally answer the long-standing debate about whether it's the same as fibromyalgia then?

Or did a fibro GWAS study also report finding the CA pain gene?

Eta sorry but I'm too severe to watch the video at the moment. @Andy Apologies I did actually try to look at the FAQs but on my mobile only a couple results ones show up and I couldn't see the answer.
All I can tell you is this from the FAQs.

"Are these findings unique to ME or have they been found in other illnesses?

The signals we have found are different from those found in other illnesses to date, except for the one on chromosome 17 that was previously found in people experiencing chronic pain."

But I can't answer as to whether sufficiently powered GWAS has been done on fibromyalgia. The GWAS my very quick look found seemed to be on very small cohorts, so I doubt we would make the claim that we have finally answered that debate.
 
Does anyone know what is the genetic vs environmental influence they’ve estimated for ME given these new GWAS results? I swear I read somewhere that it’s estimated to be only a ~10% genetic contribution? Can’t find the source again

And if it’s predominantly environment that influences ME wouldn’t a large scale EWAS study make more sense than further genetic studies?
I suspect that viewing it purely as genetic vs environmental might be to simplistic when you have complex dynamical interactions to account for which have to include stochasticity which may be neither (not genetic nor environmental).
 
GWAS tend to underestimate heritability.
I got the impression that the ME/CFS estimate still seems quite low compared to other diseases using the same method
 
I suspect that viewing it purely as genetic vs environmental might be to simplistic when you have complex dynamical interactions to account for which have to include stochasticity which may be neither (not genetic nor environmental).
Right and if true a very high predominance of non-genetic influence, whether stochastic or epigenetic, would further suggest that it’s potentially quite limited what mechanistic insights we might glean from additional genetic studies.
 
Back
Top Bottom