-
New Forum Software Installed (Still a few issues but reopening the forum) More details.
Preprint Comparing DNA Methylation Landscapes in Peripheral Blood from [ME/CFS] and Long COVID Patients, 2025, Peppercorn et al
- Thread starter Nightsong
- Start date
jnmaciuch
Senior Member (Voting Rights)
I briefly saw something about this when I was trying to find the corresponding author's email--from the first news article that came up, it looks like his family member has ME/CFS and he also has trainees with mild ME in his lab.
jnmaciuch
Senior Member (Voting Rights)
I just got a response from the corresponding author, he confirmed that figure 2 is only on significant sites. I've asked if he would mind sharing the PCA on all sites with me. He said that it was more blurry (as you'd expect) but there was still separation between groups.
Yes, I too got a response from Warren. He points out that the other senior author is a world renowned expert in this. He doesn't see any problem with the way the results have been presented with the PCA just on the 3000 or so differentially methylated fragments.
He writes that he was surprised at the differentiation between the ME/CFS and Long Covid cohorts. The manuscript goes into quite a lot of detail as to why this may be so.
But to me, I think that could easily be an artefact of the process of selecting the DMFs. Maybe a mathematician could think about this problem and tell me if I am thinking about this right?
If you take a set of DMFs for the ME/CFS group and then the set of DMFs for the LC group, some of those DMFs, possibly many of them, will just be noise that is randomly unique to each group. So then, if you do a PCA on the combined set of DMFs, it seems to me obvious that the ME/CFS group and the LC group will be separated from each other, as well as from the healthy controls.
He writes that he was surprised at the differentiation between the ME/CFS and Long Covid cohorts. The manuscript goes into quite a lot of detail as to why this may be so.
But to me, I think that could easily be an artefact of the process of selecting the DMFs. Maybe a mathematician could think about this problem and tell me if I am thinking about this right?
If you take a set of DMFs for the ME/CFS group and then the set of DMFs for the LC group, some of those DMFs, possibly many of them, will just be noise that is randomly unique to each group. So then, if you do a PCA on the combined set of DMFs, it seems to me obvious that the ME/CFS group and the LC group will be separated from each other, as well as from the healthy controls.
I find the overlap between the ME/CFS and LC group in terms of shared DMFs pretty interesting. I think if the >10% different DMFs were entirely random, possibly we'd expect to see something like 27 DMFs that were shared.
But there are 118 DMFs that were common between the two patient groups. So, to me, that's suggesting quite a lot more commonality between the two groups than we would expect from a random collection. And that's one of the problems with that PCA chart - an artefact of the way it has been constructed leads people to thinking that the two groups are more different than they actually are.
But there are 118 DMFs that were common between the two patient groups. So, to me, that's suggesting quite a lot more commonality between the two groups than we would expect from a random collection. And that's one of the problems with that PCA chart - an artefact of the way it has been constructed leads people to thinking that the two groups are more different than they actually are.
But they wouldn't be entirely random because you've already done a pre-selection from 73239 to 3363 (whilst this preselection is not based on a preselection of methylation difference but rather on statistical significance, that doesn't mean that the one can't effect the other), which might mean in your 3363 DMFs you're expecting to see more DMFs that pass the >10% threshold than you would purely out of luck in 3363 samples. I guess to know the effects of that you'd have to know how noisy this data tends to be.
Jonathan Edwards
Senior Member (Voting Rights)
Your argument sounds right to me and I am afraid that this is a mistake that very renowned experts can make. But it ay not apply in this case. I have been too busy with other things to go through it.
Did they group the ME/CFS and LC cohorts together before doing the significance test to pick 3663 DMFs that were differentially methylated compared to HC? If so, I don't immediately see an issue. Separation from the HCs on the plot would be expected, but not between LC and ME/CFS, I think.
If each group was tested for DMFs comparing to HCs separately, then even if all the DMFs were completely random, it'd make sense that different false positive DMFs would be selected in each group and would thus separate LC and ME/CFS on the plot.
My reading of the text is that the ANOVA was run on the 3 separate groups.
Oh yeah, I see that now. Then it seems to me that the separation is to be expected.
I read some other papers on methylome analysis in cancer and other diseases and it is always expected to work with significant fragments that pass a certain P value criteria or FDR criteria (in larger sample number) to avoid the noises or false fragments.
As per the manuscript in this one says, the PCA showed three different clusters which is very interesting and in discussion the authors have mentioned that the separation can be because of many things, one can be that Long Covid patients have been suffering through the disease for 1 year in this study, whereas ME patients have been for 12 years, so the difference between their methylation is obvious to be identified in the PCA.
Also even though there are similarities in both the diseases as mentioned through the 118 DMFs which is very interesting finding, there are dissimilarities as well as mentioned in the 26 DMFs. In the box plots as well in some cases I saw the methylation is going in separate directions for certain fragments in ME and LC. This paper seems to suggest that ME and LC are very similar (even with so low patient numbers) but there are obviously certain differences between them, which is eventually found in the PCA, and it will also help in future to separate people from ME and LC compared to HC. Two diseases cannot be 100% similar, and it's really good that there is separation observed, which means that methylation was actually able to separate them, showing promise in future.
I hope the authors get funded for a bigger study to find a diagnostic biomarker for both LC and ME/CFS. Peace!
As per the manuscript in this one says, the PCA showed three different clusters which is very interesting and in discussion the authors have mentioned that the separation can be because of many things, one can be that Long Covid patients have been suffering through the disease for 1 year in this study, whereas ME patients have been for 12 years, so the difference between their methylation is obvious to be identified in the PCA.
Also even though there are similarities in both the diseases as mentioned through the 118 DMFs which is very interesting finding, there are dissimilarities as well as mentioned in the 26 DMFs. In the box plots as well in some cases I saw the methylation is going in separate directions for certain fragments in ME and LC. This paper seems to suggest that ME and LC are very similar (even with so low patient numbers) but there are obviously certain differences between them, which is eventually found in the PCA, and it will also help in future to separate people from ME and LC compared to HC. Two diseases cannot be 100% similar, and it's really good that there is separation observed, which means that methylation was actually able to separate them, showing promise in future.
I hope the authors get funded for a bigger study to find a diagnostic biomarker for both LC and ME/CFS. Peace!
Last edited by a moderator:
I think there has been a mistake interpreting this, I think it says that the three authors mentioned here are funded by the ANZMES and donations from the families and other resources as well, not that the three authors part funded the study. This seems more logical though.
Yes, I'm pretty sure that they chose the DMFs based on comparisons between LC and HC, and ME/CFS and HC. So, I think the PCA is primarily an illustration of their DMF selection process rather than showing any truth about the data.
Warren also seemed to think that the consistency of the results within the groups as shown by the close clustering of the individuals in each group on the PCA chart said something real about the data. But, the researchers specifically selected fragments where the P value for the comparison between a disease cohort and the controls was good. Intuitively, that seems likely to me to produce clustering within groups.
Even if I'm right about this, does it matter? I think it does. It's a bit like the subjective outcomes in unblinded trials recipe - the process, I think, is a guaranteed way of getting a certain outcome. But, it doesn't actually mean anything, it misleads and, it affects credibility. And it leads people to make faulty conclusions e.g. about how different the LC and ME/CFS groups are.
I'd love to get hold of the data for the 70,000 odd fragments, and then reallocate the members of the matched trios across three new random groups, select DMFs in the way the researchers did (but for the new groups) and see if fairly similar PCAs could be produced. I think that they probably could be.
Warren also seemed to think that the consistency of the results within the groups as shown by the close clustering of the individuals in each group on the PCA chart said something real about the data. But, the researchers specifically selected fragments where the P value for the comparison between a disease cohort and the controls was good. Intuitively, that seems likely to me to produce clustering within groups.
Even if I'm right about this, does it matter? I think it does. It's a bit like the subjective outcomes in unblinded trials recipe - the process, I think, is a guaranteed way of getting a certain outcome. But, it doesn't actually mean anything, it misleads and, it affects credibility. And it leads people to make faulty conclusions e.g. about how different the LC and ME/CFS groups are.
I'd love to get hold of the data for the 70,000 odd fragments, and then reallocate the members of the matched trios across three new random groups, select DMFs in the way the researchers did (but for the new groups) and see if fairly similar PCAs could be produced. I think that they probably could be.
Welcome to the forum @Pikachu.
I crossposted with you, but I think your post illustrates the issue that I mentioned - the PCA leads people to make conclusions that aren't supported by the data. It's my understanding that the Long Covid patients here were essentially post-Covid-19 ME/CFS patients, so we can't assume that they have a different disease.
Do you have some involvement with this study?
I crossposted with you, but I think your post illustrates the issue that I mentioned - the PCA leads people to make conclusions that aren't supported by the data. It's my understanding that the Long Covid patients here were essentially post-Covid-19 ME/CFS patients, so we can't assume that they have a different disease.
Do you have some involvement with this study?
It
It would be interesting to get hold of the 70000 data to see what the results so for PCA, however my understanding would be that the paper is just a small study and we all are too hyped up about this. I would give the authors benefit of the doubt and let the manuscript be peer reviewed and see if the reviewers pick up something that we have been discussing about and see if they seem the paper to be deemed fit. I am interested to see more such studies from other researchers comparing Long COVID and ME/CFS. Again to be noted that in this study Long COVID patients are just suffering from the disease for 1 year where ME/CFS for 12 years so that would make them cluster separately.
It would be interesting to get hold of the 70000 data to see what the results so for PCA, however my understanding would be that the paper is just a small study and we all are too hyped up about this. I would give the authors benefit of the doubt and let the manuscript be peer reviewed and see if the reviewers pick up something that we have been discussing about and see if they seem the paper to be deemed fit. I am interested to see more such studies from other researchers comparing Long COVID and ME/CFS. Again to be noted that in this study Long COVID patients are just suffering from the disease for 1 year where ME/CFS for 12 years so that would make them cluster separately.
jnmaciuch
Senior Member (Voting Rights)
I have been able to view the PCA on the 70K sites. I don't have permission to share the image itself, but I do have permission to describe it.
As expected, much much less clean separation between groups, all the points are quite dispersed through the 2D space. PC2 gives hints of separation by group, with LC and HC opposite eachother and ME/CFS generally sandwiched in the middle (though it is a messy and imperfect trend).
The randomness of distribution along PC1, plus a bit of loose clustering in one part of the plot, leads me to think that age and/or sex are driving most of the variation. Which is to be expected, as methylation is certainly known to change with age.
For comparison, my previous work with ATAC-seq was on aging, and therefore had the benefit of the main factor of interest driving most of the variation anyways. Still, I always like to see the full PCA included as a supplemental precisely for this reason.
Dr. Tate mentioned that he would be modifying the results section to avoid giving the wrong impression [edit: about figure 2].
As expected, much much less clean separation between groups, all the points are quite dispersed through the 2D space. PC2 gives hints of separation by group, with LC and HC opposite eachother and ME/CFS generally sandwiched in the middle (though it is a messy and imperfect trend).
The randomness of distribution along PC1, plus a bit of loose clustering in one part of the plot, leads me to think that age and/or sex are driving most of the variation. Which is to be expected, as methylation is certainly known to change with age.
For comparison, my previous work with ATAC-seq was on aging, and therefore had the benefit of the main factor of interest driving most of the variation anyways. Still, I always like to see the full PCA included as a supplemental precisely for this reason.
Dr. Tate mentioned that he would be modifying the results section to avoid giving the wrong impression [edit: about figure 2].
Last edited:
Thank you for the warm welcome. I couldn’t find the thing you mentioned that the LC patients are Post COVID 19 ME/CFS patients, as to my understanding from the paper, the Long COVID patients are different and never had ME before, but obviously we can ask Dr. Tate for a confirmation?
i have a few friends who is suffering from ME and few from Long COVID, so I came across this paper while looking for the disease as this is the most recent paper. Unfortunately I do not have any contributions to the paper.
Utsikt
Senior Member (Voting Rights)
Personally, I don’t put too much value on the peer review process in general. Far too much rubbish gets through. At this point, it has become a meme at the same level of the «gold standard» RCTs.
If there is one thing many of us here have learned over the years, Pikachu, it's that what is generally assumed to be true may not be. Subsets of that are skepticism about the beliefs that 'world renowned experts are always right' and 'the peer review system ensures only quality research gets published'.
These researchers will hopefully go on to produce more ME/CFS papers using this technique, perhaps more influential ones. Even a small paper like this can affect how people think about the question of ME/CFS (and Long Covid). Here we are trying to improve the quality of research, so that answers about this disease are found sooner, rather than later. If scarce research funds are used, we want them to be applied in the best way possible.
The description of the criteria for inclusion in the LC cohort is a bit vague, but 'symptoms indicating an ME/CFS-like syndrome' suggests that the LC people look much like the ME/CFS people in terms of symptoms.Methods said:
jnmaciuch
Senior Member (Voting Rights)
The methods mention that they used the WHO's 2021 case definition, which says this:
So the described features are ME/CFS-like, certainly, if you do not take into account degree of severity or post exertional malaise (considered to be the hallmark characteristic of ME/CFS, if you are unfamiliar).
From the text it does not seem that CCC criteria for ME/CFS were assessed for the LC cohort, though we know from other studies that many LC patients do meet ME/CFS criteria (though the actual proportion who do is not possible to know due to unreliable categorization in previous studies).
[Edit: oop, cross-posted with @Hutan]