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

[Hoopoe]

I got the impression that the ME/CFS estimate still seems quite low compared to other diseases using the same method

UKB SNP-Heritability Browser

nealelab.github.io

Comparison of Family History and SNPs for Predicting Risk of Complex Disease - PMC

The clinical utility of family history and genetic tests is generally well understood for simple Mendelian disorders and rare subforms of complex diseases that are directly attributable to highly penetrant genetic variants. However, little is ...

pmc.ncbi.nlm.nih.gov

If you read the Wikipedia article, it explains that the more SNPs are considered, the more accurate heritability estimates from GWAS become.

Things like body parameters have presumably heritability estimates that can draw on far more SNPs than diseases.

I'm not sure that the second article is relevant.

 

[bobbler]

Previously I wondered if the “Chronic pain gene“ (CA10 I think?), might actually be associated with ME not chronic pain. But since ME is commonly undiagnosed and has chronic pain for many, a lot of pwME may have ended up in a chronic pain GWAS and with large sample sizes it might show an effect.

I'm going to try and now find the thread on this specific gene. As I'm thinking is it pain or chronic pain, and which things does it relate to etc

Wary of getting the terminology right on this because some of the chronic pain sector, and eg one of the specific guidelines for something called eg chronic pain was a problem / quite bps wasn't it?

it is a term that currently is a bit of a bucket/one of those that has a variety of layman's meanings but also covers quite specific diagnoses, where I don't know what the finding it in the GWAS specifically it is referring to

 

[ME/CFS Science Blog]

If you read the Wikipedia article, it explains that the more SNPs are considered, the more accurate heritability estimates from GWAS become.

Things like body parameters have presumably heritability estimates that can draw on far more SNPs than diseases.

To clarify: the link to the Neal lab data includes a search function where you can find the heritability estimate for other diseases using UKB methods similar to what DecodeME used. It seems that ME/CFS isn't like schizophrenia, Crohn's disease or diabetes type 1 for example, which have a significant higher heritability.

Unless the rare variants shows up very interesting things, I do think this points to relatively low heritability of ME/CFS?

 

[Amw66]

Genes also multitask .
So for feedback loops etc it may not be the more obvious association / function which may be relevant

 

[Jonathan Edwards]

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

My memory is that a US- based twin study suggested a much higher genetic component - maybe 40-50%. I think Chris was disappointed that the SNPs only provided evidence for a modest ~10%.

In view of the fact that the obvious suspects like MHC genes did not turn up any major risk effect over a period of twenty years, together with the elusive nature of the ME/CFS pathology, I don't think it is that surprising that the score so far is lower than a lot of other diseases.

But I don't think this is a major reason to draw back from genetics. In other diseases rare gene variants that provide high risk for disease in a tiny proportion of cases can be extremely useful. The rare cases of homozygous complement gene defects show that one sure fire way to get lupus is through defective complement. Which why a whole genome sequencing project makes sense to me. And I presume that you would be lucky to get much out of it with less than 5,000 cases, just because the search is for rare gene variants.

 

[Utsikt]

Right but the underlying goal of a genetic study like this is to gain some mechanistic insights as to what triggers or continues to drive the illness, and if the genetic component/influence is very low then these gene variants more likely than not do not play a major role in the pathomechanism and therefore their utility in further research and treatment development is rather limited

I’m not sure that’s the right interpretation. If the genes show up, they are always causal, so they do play a role in the disease mechanism(s).

 

[Utsikt]

Do we know the heritability and the explanatory power of genetics for other diseases that are believed to primarily have an infectious onset?

 

[Jonathan Edwards]

Unless the rare variants shows up very interesting things, I do think this points to relatively low heritability of ME/CFS?

I am not sure that you can conclude that even after chasing rare genes. If genetic risk comes from combinations of genes looking at genes one at a time is not enough. That may be a rare situation but it something that cropped up in the context of NK receptors for Class I. We are all born with six Class I alleles and a whole bunch of receptor alleles that are supposed to fit the Class I. A disease might be linked to certain pairs of alleles or even more complex combinations.

 

[Hoopoe]

To clarify: the link to the Neal lab data includes a search function where you can find the heritability estimate for other diseases using UKB methods similar to what DecodeME used. It seems that ME/CFS isn't like schizophrenia, Crohn's disease or diabetes type 1 for example, which have a significant higher heritability.

Unless the rare variants shows up very interesting things, I do think this points to relatively low heritability of ME/CFS?

Okay, I think I understand you better. You were comparing heritability of ME/CFS to other diseases. The liability heritability of Crohn's using LDSC is around 24%. That's higher than the 10% found by DecodeME. I find that disappointing as well. Maybe the estimate can be revised substantially in the future.

I was comparing the 10% heritability estimated by DecodeME with other ME/CFS studies that used different approaches and found much higher heritability.

 

[ME/CFS Science Blog]

My memory is that a US- based twin study suggested a much higher genetic component - maybe 40-50%

I think it was this study:
Repurposing large health insurance claims data to estimate genetic and environmental contributions in 560 phenotypes (2019) Lakhani et al. | Science for ME

Here's the overview from Dibble et al.

Of three studies that have estimated narrow-senseheritability (h2) using large cohorts, two reported non-zero h2-values that provide evidence for heritability of risk for CFS and,presumably, ME/CFS. An analysis of US health insurance claimeda high narrow-sense heritability (h2 = 0.48) of CFS (23), whereasan analysis of the UK Biobank individuals self-reporting a CFSdiagnosis reported a less striking heritability (single nucleotidepolymorphism- [SNP-] based approximate h2 = 0.08 with low confidence) (24) (http://www.nealelab.is/uk-biobank). The third, alarge twin-based study of CFS-like cases, produced an inconclusive result, with the 95% confidence interval of h2 including zero[0.03 (0.00–0.65)] (25)

 

[Hoopoe]

This study claims estimates of heritability using LDSC underestimate by 30% (and 50% in simulations) https://pmc.ncbi.nlm.nih.gov/articles/PMC10066905/

 

[ME/CFS Science Blog]

The liability heritability of Crohn's using LDSC is around 24%. That's higher than the 10% found by DecodeME. I find that disappointing as well.

I kind of picked examples with high heritability though. There are other diseases with similarly low heritability estimates such as rheumatoid arthritis so the estimate for ME/CFS is nothing weird. But perhaps a bit lower than some expected.

 

[Yann04]

I’m not sure that’s the right interpretation. If the genes show up, they are always causal, so they do play a role in the disease mechanism(s).

(assuming perfect methodology including selection which in practice is impossible, there will always be some caveats).

But yes, it is very likely casual.

 

[Jonathan Edwards]

Do we know the heritability and the explanatory power of genetics for other diseases that are believed to primarily have an infectious onset?

Reiter's is the obvious one. We knew that there was an arthritis syndrome that occurred after dysentery or non-specific urethritis. It was discovered that a high proportion of patients are B27+, like ank spond and various other things allowed us to build a model of disease around the whole lot. Reiter's has a slightly different risk profile though if I remember rightly - B7 also confers risk. We are still waiting for Matt Brown to work out why these HLA-B alleles are important but one day we will know.

I think rheumatic fever is predisposed to by some immune response genes too but I forget which.

 

[Hoopoe]

ME/CFS does have a problem with misdiagnosis, and my intuition tells me that heritability estimates will be more accurate for diseases where there's less diagnostic uncertainty.

In the Neale database, h2 for fibromyalgia is 0.0112 (or about 1%).

ME/CFS is doing pretty well in comparison.

 

[Hutan]

I've been wondering about the lack of sex chromosome results in the preprint. Chris said in the interview with David that they had results by last September, October. And surely, for a disease where more women are diagnosed than men, the X chromosome would be the first one you would look at closely, not the last?

It's probably been a year since the DecodeME team started stressing that ME/CFS is a female disease, and people started to look at the evidence for the "twin peaks" idea for females onset. DecodeME publications have referred to a sex ratio of 4 to 1, which I think is higher than the evidence really warrants, and doesn't take into account the considerable bias against men choosing and being given an ME/CFS label.

There's been the odd comment about 'if there's not a genetic difference between males and females, it must be hormones making the difference'. And we've had OMF pushing for a study on hormones, stretching the evidence in order to justify one.

So, it's a bit puzzling. I'm assuming the sex chromosome results weren't prioritised for inclusion in the initial preprint for a strategic reason. I'm not sure what the reason is. @Andy, Simon M?

 

[ME/CFS Science Blog]

DecodeME publications have referred to a sex ratio of 4 to 1, which I think is higher than the evidence really warrants

The DecodeME participants were 85% females themselves.

So, it's a bit puzzling. I'm assuming the sex chromosome results weren't prioritised for inclusion in the initial preprint for a strategic reason. I'm not sure what the reason is.

I think it's the default of GWAS because sex chromosomes have additional difficulties to analyze see:
eXclusionarY: 10 years later, where are the sex chromosomes in GWASs? - PMC

EDIT: added quote from the paper above

The X chromosome presents multiple analytical challenges,9,10,11,12,13,14 including (1) a male has one copy of the X chromosome while a female has two, in contrast to the autosomes; (2) the X chromosome in male germ cells only recombines with the Y chromosome in the pseudo-autosomal regions (PARs) but not in the NPR; (3) in contrast to males, the two copies in female germ cells recombine across the entire X chromosome; (4) the two female copies are also subject to X inactivation (i.e., X chromosome dosage compensation); (5) the X-inactivation status at the population level can be random, skewed, or absent (i.e., X-inactivation escape); and (6) the true X-inactivation status at the individual level cannot be derived from GWAS data alone.

 

[leokitten]

I’m not sure that’s the right interpretation. If the genes show up, they are always causal, so they do play a role in the disease mechanism(s).

From my understanding GWAS findings by themselves aren’t causal only correlative. You need to perform additional analyses to do casual inference. See for example an explanation here https://pmc.ncbi.nlm.nih.gov/articles/PMC7614231/

GWAS estimate the association of each SNP with the phenotype, not the causal effect of each SNP on the phenotype.

 

[Jonathan Edwards]

I've been wondering quite a bit about the lack of sex chromosome results in the preprint. Chris said in the interview with David that they had results by last September, October. And surely, for a disease where more women are diagnosed than men, the X chromosome would be the first one you would look at closely, not the last?

It's an interesting question but I am not sure it is obvious to look at the X and Y chromosomes. The reason why women are more likely to get ME/CFS than men is that they have two X chromosomes and men have one plus a Y. We know that already. We are not expecting women with ME/CFS to have any different gene variants from other women, or indeed X gene variants from men.

Maybe we can argue that if we find women with ME/CFS have a higher rate of a variant of one gene that partly controls the difference in immune response between men and women and not another that does in a different way then the first way is more relevant. But I am not sure we even know which genes individually control the differences, or how we could test for that other than by complicated gene editing experiments on the X chromosome (and it would have to be mice). And these genes may not be polymorphic anyway.

 

[Jonathan Edwards]

From my understanding GWAS findings by themselves aren’t causal only correlative. You need to perform additional analyses to do casual inference. See for example an explanation here

I think you need to do other things to establish the precise nature of the causation and maybe the strength but if you correctly identify a gene polymorphism as linked to risk and have excluded linkage disequilibrium (assuming you can) then you have established a causal relation. That is the beauty of genetic studies.