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

[ChronicallyOverIt]

Right at the moment we are unable to give an estimate on that. There are further analyses yet to be completed and the team are blocked from working on them due to the UK Biobank access issues which are explained upthread.

Gotcha. Thanks for the information! Looking forward to that!

Maybe just point out that it is considered good enough for the UK government to provide another ~£5M on the basis of this work to do more. There are times when raising peer review tends to look like an excuse for not wanting to believe in something!

Yes they were definitely being extra pedantic and not engaging in good faith

 

[DMissa]

I really appreciate how you’re paying so close attention to everything that’s going on, and being willing to adjust the course along the way! We need more people that appreciate the value of null results.

well i'm glad it's at least noticeable.. if i wasnt making useful contributions on the rare opportunity that i get to design a project myself from the ground up i'd rather miss out on the funds and be forced to leave academia

 

[Turtle]

well i'm glad it's at least noticeable.. if i wasnt making useful contributions on the rare opportunity that i get to design a project myself from the ground up i'd rather miss out on the funds and be forced to leave academia

More than noticeble, highly appreciated even the things that are way out of my league.

I have a wild but layperson's idea, have a good laugh when it is nonsense and disregard or think about it.
I once heard Maureen Hanson on a conference mention in a bit of a casual way that the polarity of the mitochondria had changed. To me that sounds huge! What if phosfolipids don't know whether it has to be heads or tails in the polarity confusion?

 

[DMissa]

I once heard Maureen Hanson on a conference mention in a bit of a casual way that the polarity of the mitochondria had changed. To me that sounds huge! What if phosfolipids don't know whether it has to be heads or tails in the polarity confusion?

If I could guess this may be from Alex Mandarano or Jessica Maya's T cell experiments which (from memory...) showed reduced mean values that describe mitochondrial membrane potential ("MMP") in some T cell subsets. I can't think of other evidence from their lab that I am aware of that would prompt use of the word polarity with relation to mitochondria.

Mitochondria are bound by a double membrane envelope. There is machinery on the inner membrane that pumps positively charged particles into the intermembrane space (the gap between the two membranes). This creates an electrochemical charge build-up or "potential". This is the MMP. The MMP is consumed mainly by an enzyme complex that produces ATP. This is the "final" important step of "respiration" which is what is taught in high school as the thing that makes most of your body's energy.

I have no idea about what this means for the illness or anything to do with potentially aberrant lipid biology. In my view if there is something seen to be different in energy generating processes in an immune cell type it probably represents a difference in something relating to their maturation or activation status, as the field of immunometabolism shows time and time again that very specific subsets are highly dependent on fine tuned changes to these processes. This is not necessarily the explanation but would be the one I would favour as most likely in a vacuum.

If these sorts of observations (including things I've published in far more naive times) did reflect an underlying energy production issue my questions would be: why isn't it seen in the other cell types measured (why isn't it systemic, what's causing it?) and why doesn't ME/CFS involve the stereotypical features of energy deficiency disorders such as wasting, developmental/morphological abnormalities when onset is pediatric, inexplicable hypertonia, etc?

If it is a relevant part of pathology then finding the cause and consequences of it are important. MMP is a steady state. There are things that build it up and things that deplete it. A shift in either or a relative imbalance in both can lead to a shift in the steady state. Alone these measurements do not say which possibility is the case.

I do not think this is altogether relevant to the genetic evidence under discussion in this particular thread but I thought worth discussing in reply to your comment. Also, don't put yourself down, there is nothing silly about your question. Scientists are just as human as you are and biology is too large for one person to know most of how it works. I hope everything I've written makes sense for you.

 

[ME/CFS Science Blog]

Was reading this 2025 paper with whole-genome sequencing data from more than 300.000 people in the UK Biobank. Adding confounders had little effect on the heritability estimates for most traits, except for educational attainment and fluid intelligence.

However, uncorrected estimates of hWGS2 for educational attainment and fluid intelligence score were significantly inflated by fine-scale geographical structures in the United Kingdom that were not fully captured by genotypic principal components. This underscores the importance of using geographical information to inform and correct biases affecting heritability estimates19,20, especially for behavioural traits involved in migration patterns

Source: Estimation and mapping of the missing heritability of human phenotypes | Nature

Given the concern that the DecodeME results might reflect a bias like this, I wonder if it would be an option to add more potential confounders (i.e. more PCs) to see how it affects the results.

 

[Chandelier]

Bluesky post by Chris Ponting.
One year ago the preprint has been published!

Chris Ponting
6. August 2026
Today is the 1-year anniversary of the #DecodeME genetics preprint.
It was an emotional day for many including everyone in the team that delivered the project #pwME #MEcfs @actionforme.bsky.social institute-genetics-cancer.ed.ac.uk

https://institute-genetics-cancer.ed.ac.uk/sites/default/files/2026-05/2025-08-03 DecodeME Preprint.pdf

institute-genetics-cancer.ed.ac.uk

Since then, we & others have been working hard to better understand these results.
With replicated results, @precisionlife.bsky.social found that ME/CFS is a complex genetics condition link.springer.com

Identification of novel reproducible combinatorial genetic risk factors for myalgic encephalomyelitis in the DecodeME patient cohort and commonalities with long COVID - Journal of Translational Medici...

link.springer.com

An analyst found that ME/CFS genetic risk is especially concentrated in neuronal subsets trafalmadorian97.github.io

MAGMA HBA (DecodeME, ME/CFS) - ME/CFS Bioinformatics Home

trafalmadorian97.github.io

But, this year, as we were preparing an update, the UK Biobank analysis platform shut us & everyone out.
It’s due to reopen in September but with limited functionality.
So realistically the new update will now be in 2027.

Looking back, the DecodeME project’s primary importance is its objective and statistically/technically robust evidence that ME/CFS is an organic disease.
The alternative BPS hypothesis that it’s not a disease but an illness belief has, by contrast, no similarly strong evidence.

ME/CFS - with Long Covid - should now be treated by Governments and society as a common and highly debilitating disease.
The neglect, scorn and gaslighting must stop.Meanwhile, we keep going sequencing the entire genomes of 6,000 #pwME in the #SequenceME and Long Covid project.
actionforme.org.ukinstitute-genetics-cancer.ed.ac.uk

Sequence ME & Long Covid

www.actionforme.org.uk

 

[ME/CFS Science Blog]

Was wondering if there's an explanation for why the LDSC intercept for DecodeME is below 1.

Higher than 1 is usually seen as a sign of bias such as population stratification but below 1 is less clear. Could it mean that part of the true signal was taken away by the principal components?

 

[forestglip]

Was wondering if there's an explanation for why the LDSC intercept for DecodeME is below 1.

Higher than 1 is usually seen as a sign of bias such as population stratification but below 1 is less clear. Could it mean that part of the true signal was taken away by the principal components?

I don't fully understand it, but this might be helpful:

https://github.com/bulik/ldsc/wiki/FAQ

Q. Why is my single-trait LD Score regression intercept below one?

A. If the LD Score regression intercept is non-significantly less than one, If your summary statistics were generated from GC corrected data (even data that were only single GC corrected), the intercept should be less than one (see the supp note of Bulik-Sullivan et al., Nature Genetics, 2015). If your summary statistcs contain chi2 statistics for SNPs with low minor allele count, then the chi2 statistics for these SNPs may be deflated relative to the asymptotic distribution. Try filtering out low MAF SNPs. If mean chi2. is below one, ldsc will not work properly.

 

[ME/CFS Science Blog]

I don't fully understand it, but this might be helpful:

Thanks!

My guess is that genomic control (GC corrected) refers to dividing by the genomic inflation factor: the ratio of your median chi square value by 0.4549 (what the median chi square value would be under the null with no bias). This was an (older) approach intended to control for inflation due to bias but it wasn't applied to the DecodeME summary data so don't think this is the reason.

 

[ME/CFS Science Blog]

Don't understand everything, but there's some discussion that eQTL data and GWAS hits often do not match very well. Genes that are likely to be causally related to disease often do not have a lot of eQTL data.

Found some more support for this in this preprint (Ji et al. 2025)
Benchmarking genome-wide association study causal gene prioritization for drug discovery

They looked at genetic evidence leading to successful drug development. Distance (the nearest gene method) did equally well compared to complex machine learning methods like L2G. eQTL data actually made the prediction worse.

We found that the L2G score (OR = 3.14, 95% CI of 2.31 to 4.28; Figure 2a) and the nearest gene method (OR = 3.08, 95% CI of 2.25 to 4.11) were similarly predictive of drug success, while eQTL colocalization did not significantly predict drug approval (OR = 1.61, 95% CI of 0.92 to 2.83)

Finally, we tested whether eQTL colocalization or L2G provide additional value for nearest genes, which as noted above are associated with greater odds of drug approval (OR = 3.04, 95% CI of 2.25 to 4.11; Figure S1). L2G performed similarly to the nearest-gene heuristic alone (OR = 3.27, 95% CI 2.39 to 4.49). In contrast, eQTL colocalization was associated with a lower odds ratio (OR = 2.29, 95% CI 1.26 to 4.17). This suggests that within the set of nearest genes, those with additional eQTL support are less likely to be approved than those without, mirroring the negative trend observed when taking out nearest genes.

 

[ScoutB]

Very interesting. Taking a glance at forestglip's table it sort of looks like a lot of the genes that we've found interesting leads on are indeed the nearest ones.

examples of nearest genes that seem a bit interesting:
HTT, DCC, OLFM4 turned up in fibro (so did tier 1 gene RABGAP1L to be fair)

DCC and CACNA1E were mentioned in that interesting study on nerve pain SNT posted.

SLC2A14 was found to be differentially expressed in progenitor cells from pwME/CFS here.

Again DCC and OLFM4, this time in adolescent-onset depression

Edit: SHISA6 has all kinds of connections listed on its thread.

Of course tier 1 genes like CA10 are very promising too. Have not performed an unbiased scan