What would successful brainstorming about ME/CFS genes look like?

Great start @Sasha! Usually I’ll start at gene cards as well, and then the first thing I’ll do is scroll allllll the way down to the tissue expression chart that looks something like this:

IMG_9784.png
That’s a compilation from various databases looking at where RNA for that gene has been detected in the human body. Detected RNA doesn’t always mean that the protein is being produced, but it can be a good starting point for understanding where it might be playing important roles. The Human Protein Atlas is also a good resource, as it can sometimes tell you where the protein has actually been detected (i.e. where the gene is actually functioning). That will give you a good idea of whether it’s a gene that is really only functioning in one organ/context, or if it’s a gene that has just been characterized extensively in only one context.

The summary sections of gene cards are usually a mishmash of one-line summaries from various papers that mention the gene name is some context, so read them as such. There is no coherent narrative being presented here. Unfortunately gene cards is not particularly forthcoming about the sources that these summaries are drawing from. If something sticks out to me, I’ll usually just Google the terms and the gene name and see what comes up, like “ERK pathway SYNGAP1” or “Autism SYNGAP1”.

Usually studies related to those more technical terms will be cell biology studies where they knocked out the gene and found that the cell was unable to do XYZ. Sometimes you’ll find knockout mouse models for exploring more complicated organ systems. The ones related to human disease names are going to be genetics studies that provide no mechanistic insight. If the paper is pure technical gobbledygook you can skip to the discussion, just keeping in mind that you’re reading someone’s attempt to weave a narrative and it may present a shinier story than what the data actually shows.

I agree with @hotblack that AI might be good for defining basic concepts—I’d expect it to at least point you in the right direction for what the ERK pathway is, for example. But if you ask it “how might SYNGAP1 relate to ME/CFS?” it will just draw from whatever half baked theories people have proposed about HPA axis or neural connectivity, probably.

Once you do this for enough genes you’ll probably see certain words and phrases start to come up that paint a story in your head. And of course that process of pattern recognition will often be driven by confirmation bias, but there’s not much you can do about it for these purposes. It’s still a useful exercise just for driving some learning!
 
Another good starting point is the ME/CFS paper you get a gene from in the first place.

The authors are likely to have done at least a scan of the literature to see how the significant genes might fit into the explanation for ME/CFS. So they did a bit of the work for you, and you can check if the things they linked make sense and continue researching from that starting idea.

For example, from the Zhang paper:
As highlighted in our network analysis, ME/CFS genes participate in biological pathways associated with synaptic function (M20; Fig. 4B and 4D). SYNGAP1, another gene driving downregulation of the ME/CFS genes in cytotoxic CD4 T cells (Fig. 4G) and a member of M20, is involved in synaptic signaling and plasticity, essential for brain function with mutations linked to neurodevelopmental and psychiatric disorders53. SYNGAP1’s role in synaptic signaling highlights its potential connection to neurological symptoms in ME/CFS, offering therapeutic potential in neuroprotective strategies54.
Click to expand...

And they link to a couple papers that they base these claims on (the 53 and 54).
 
Last edited:
Jonathan Edwards said:
Except that the great thing about gene linkages is that they are always the root cause.
Click to expand...
Isn't one of the problems going to be that genes are sometimes inherited in strings of genes rather than as individual units, so if you have a disease association with a particular SNP, it might not be the gene containing that SNP wot dunnit? But the gene next to it, or a bit further down the line?

Do adjacent genes tend to produce proteins that do related stuff, such as affect sleep, for instance?
 
Sasha said:
But the gene next to it, or a bit further down the line?
Click to expand...

You're getting a bit technical here, @Sasha. This is what imputation protocols are for.God knows how they work.

Sasha said:
Do adjacent genes tend to produce proteins that do related stuff, such as affect sleep, for instance?
Click to expand...

Sometimes, sometimes not. For instance, in the MHC you have Class I genes at one end, whose proteins bind to CD8 T cells, and Class II genes at the other, whose genes bind to CD4 T cells and Class III genes (someone boobed here) in the middle that do all sorts of unrelated jobs making cytokines, complement and whatever. It is a bit like a cutlery draw with knives on the right, forks on the left, and the middle full of corkscrews, garlic crushers and rubber bands.

Lots of genes evolved by one gene getting duplicated, maybe through a slightly off target chromosome arm switching in a germline cell, and then the duplicate mutating to do a similar but slightly different job from the original. That makes sense if you want a whole row of receptors all with the same anchor tail and basic shape but each with a different 'receiving' end tofit something different.
 
forestglip said:
Any chance you can explain the difference in the three columns and why they're not the same? (RNAseq, SAGE, and microarray)
Click to expand...
It’s just three different methods to detect RNA from tissue—there are technical variations in how the RNA is captured, whether you need to know the sequence ahead of time or if it’s an unbiased discovery, how prone they are to issues that skew exact quantification of transcripts, the range of total transcripts that can be reliably detected etc.

SAGE and microarray are two older methods that were the precursor to high-throughput RNA-seq.

Unfortunately there’s no hard and fast rule for which one should be considered “ground truth” above the others when they disagree, so I usually try to look at what they're all telling me cumulatively.

Some of the discordance will also be because one method just didn’t have data available for that gene/tissue. That’s the most likely explanation for cases where one or two methods show high expression and another shows nothing at all
 
Last edited:
It has been suggested that ME/CFS involves genes responsible for neurodevelopment (shared genetic risk factors with autism).

It has also been reported that age of onset of ME/CFS has a peak around 15 years. I'm curious about the relationship between this onset peak and the genetic risk factors.
  • Is it possible that this is the age where the brain enters a stage of development where these identified genes play an especially important role? In other words, is the switch that turns on ME/CFS, these genes being strongly expressed at this stage of brain development?
  • Is it possible that atypical neurodevelopment occurs much earlier and that the peak in onset of ME/CFS around that age is due to a combination of other factors?
  • Maybe these genes don't increase the risk of ME/CFS by affecting neurodevelopment, but in other ways. How would we know if this is the case?
Maybe what we need is a study of gene expression changes over time in adolescents, to see if there are any patterns that stand out which are associated with ME/CFS. This sounds like it would be difficult and costly but maybe there's a way to get a sense of what's going on at this level.
 
Last edited:
Thinking about this more widely now we have some genes, is it correct to say we’re not looking for a mechanism which requires all of these genetic differences to be present but a mechanism which each of these genetic differences makes more likely to occur? Or is it even broader than that?
 
Just a mechanism that each genetic difference makes more or less liely to occur.
So a genetic difference that skews gamma delta T cells does not mean that ME/CFS has to be triggered through these cells. It just means that one route may go that way.

And there is no need for any of the genetic diferences to be present if they merely shift a threshold that a random event has to push you over. Some people with RA have none of the risk genes for RA.
 
Something that might deserve its own thread and might merit getting some further information on from Chris and his group is the list of genes that came out blank. HLA-DR does not come up and nor do any genes involved in heritable connective tissue disorders like EDS as far as I can see.

That might depend on whether the GWAS method was sensitive to allelic variations in these genes. Clearly it should be for HLA-DR but for some genes there might be blind spots perhaps.

i am beginning to think that, for instance, the absence of any links to cytokine receptor polymorphisms may be important. If symptoms are mediated by interleukins, TNF or interferons then why aren't receptor genes influencing presentation of the disease? And if these cytokines are involved in perpetuating disease through feedback loops the argument is even stronger.
 
Jonathan Edwards said:
i am beginning to think that, for instance, the absence of any links to cytokine receptor polymorphisms may be important. If symptoms are mediated by interleukins, TNF or interferons then why aren't receptor genes influencing presentation of the disease? And if these cytokines are involved in perpetuating disease through feedback loops the argument is even stronger.
Click to expand...
Isn't it more just indicating that TNF might have nothing to do with ME/CFS? Or is that what you're saying?
 
forestglip said:
Isn't it more just indicating that TNF might have nothing to do with ME/CFS? Or is that what you're saying?
Click to expand...

Yes, that is what I am saying.
We would need to be very careful but it may be that we can build a prima facie case for the following having no specific role in ME/CFS:

TNF
IFN alpha beta gamma
IL-6
IL-17
Collagen fibre integrity
Mast cells and tryptase
Specific antigen presentation on Class I and II
Various mitochondrial enzymes
And so on

And yet the positive gene linkages show that something has a specific role, with maybe a hint that that may include neuronal events and perhaps resident supportive cell housekeeping events.

A correct hypothesis will fit every bit of data you can throw at it. If we have data on lack of association with 25,000 genes, several of which we might have thought had roles, that should not be wasted.
 
Would it be worth checking the statistical significance of say, all cytokine related genes? To get an understanding of if they were just below the threshold or nowhere near? Would that be useful info or even feasible?

What about rare genes, how do they fit i to this picture? What do we need full genome analysis to be more certain of with regard to this?
 
Jonathan Edwards said:
Yes, that is what I am saying.
We would need to be very careful but it may be that we can build a prima facie case for the following having no specific role in ME/CFS:

TNF
IFN alpha beta gamma
IL-6
IL-17
Collagen fibre integrity
Mast cells and tryptase
Specific antigen presentation on Class I and II
Various mitochondrial enzymes
And so on

And yet the positive gene linkages show that something has a specific role, with maybe a hint that that may include neuronal events and perhaps resident supportive cell housekeeping events.

A correct hypothesis will fit every bit of data you can throw at it. If we have data on lack of association with 25,000 genes, several of which we might have thought had roles, that should not be wasted.
Click to expand...
Oh this is interesting. So you think the DecodeME results might have put paid to your hypothesis? Would we have definitely seen genes related to IFNG if it was involved, for example?

Are T cells 'safe' from this thought exercise or would they have been likely to have had genetic 'hits' as well? Or is HLA linked to T cells? And the HLA result is unclear isnt it... this all makes my head spin!
 
I loved the “Sherlock Holmes” approach of the hypothesis paper. It eliminates everything ME/CFS definitely isn’t, and gradually the shape of what it must be starts to emerge. Before this year the story of 99% of ME research was overattentiveness (sensitisation?!) to pseudo-positives which were statistical noise or deliberate distortions, and failure to learn from null results.

If I understand you correctly @Jonathan Edwards you’re taking - cautiously - a guesstimating approach of saying that because none of the SNPs known to be linked to a particular immune function have shown up in DecodeME, we can hypothesise that the associated function is not a significant cause of the disease. But as you say we need to treat that only as a “prima facie case” - or perhaps a prima facie alibi.

Sorry to be asking for the introductory genetics lesson, but what would a statistically robust approach look like for identifying which immune cells / functions appear not to be implicated?

There must be established approaches for calculating how big the dataset needs to be before we can say that if there are no SNPs for a particular function overrepresented in the patient dataset then it is very likely that the function is not a cause of the disease, based on the number of SNPs being considered and their frequency in the wider population? And at the level of individual SNPs, does the DecodeME dataset make it easy to identify reliably irrelevant SNPs, or will this require further analysis?
 
Jonathan Edwards said:
Yes, that is what I am saying.
We would need to be very careful but it may be that we can build a prima facie case for the following having no specific role in ME/CFS:

TNF
IFN alpha beta gamma
IL-6
IL-17
Collagen fibre integrity
Mast cells and tryptase
Specific antigen presentation on Class I and II
Various mitochondrial enzymes
And so on

And yet the positive gene linkages show that something has a specific role, with maybe a hint that that may include neuronal events and perhaps resident supportive cell housekeeping events.

A correct hypothesis will fit every bit of data you can throw at it. If we have data on lack of association with 25,000 genes, several of which we might have thought had roles, that should not be wasted.
Click to expand...
That's an ill-advised leap to make, precisely because of the limited set of genes that we know are related to those pathways. The main example would be interferonopathies, which are highly associated with genes that we know to be central in interferon signaling and regulation--the interferon receptors, IRF3/7, USP18, proteasome, etc.

But there are plenty of genes involved in interferon regulation that are a few more degrees removed. Interferon is obviously relevant in other conditions like psoriasis, and yet you don't see strong associations with interferon receptors and USP18--likely because if someone had a deleterious mutation in USP18, they'd be fast tracked towards developing a serious interferonopathy rather than psoriasis. For psoriasis, IRF2 mutations come up, which is a transcriptional regulator that plays a role in keeping constitutive interferon production in the tissue under control.

And the reality is that there are many more genes that exert regulatory effects on these pathways, but their degree of influence is far enough removed that we just simply haven't studied them in that context. So you can't say that the pathway isn't involved, you can only say that the disease doesn't involve catastrophic genetic dysregulation in the central known players of that pathway.

There may well be hits relevant to those pathways among the top DecodeME genes, we just don't recognize them as such because their role hasn't been elucidated yet. And it just might be a situation where if the mutations were any more directly involved in interferon/TNF/mitochondrial function/what-have-you, we'd be ending up with psoriasis or mitochondrial myopathies instead of ME/CFS.
 
You must log in or register to reply here.
Back
Top Bottom