A study has analysed existing genetic data in a new way to link 14 genes to ME/CFS and identify many patient subgroups. If the new approach pans out, it could transform ME research and turbocharge the development of treatments.

Paper: Genetic Risk Factors for ME/CFS Identified Using Combinatorial Analysis

Authors: Sayoni Das, Krystyna Taylor, James Kozubek, Jason Sardell, Steve Gardner

The paper has been submitted to a scientific journal and is being considered for publication. For now, the submitted draft is available as a preprint. Its findings were presented at the ME Genetics Research Summit on 14th September 2022

The study comes from Oxford-based tech company PrecisionLife. It aims to find better treatments for chronic illnesses that have few or no treatment options – such as ME.

PrecisionLife uses a technique called combinatorial analysis. Big DNA studies look for differences in single DNA’ letters’, called single nucleotide polymorphisms or SNPs, pronounced “snips”. But PrecisionLife looks for combinations of these differences. They call these combinations disease signatures.

The study looked at DNA data from nearly 2,400 people in the UK Biobank who reported in a questionnaire that a doctor had diagnosed them with ME or CFS. The analysis found 84 statistically significant disease signatures. Each was a combination of three to five SNPs, and 199 different SNPs were involved altogether.

Of the 199, the researchers focused on 25 critical SNPs that appeared in many different disease signatures. The research team used the critical SNPs to identify14 genes connected with ME.

To put this in perspective, the only previous genetic link found to ME is for an immune-system gene, a finding that, like these new ones, needs replication.

The 14 genes affect (amongst many things) energy metabolism, susceptibility to viruses and bacteria, and sleep – all of which have an obvious link to ME.

The subgroup problem

Crucially, the study looked at how the disease signatures were shared. Many disease signatures overlapped, and the researchers combined the disease signatures into 15 subgroups. The subgroups ranged in size from 5% to 30% of the biobank ME sample. 91% of patients fell into a subgroup.

This is consistent with the belief of most researchers that ME/CFS is a mix of many different subgroups of patients. Each subgroup could be a different subtype of disease or even a completely different disease. This makes it very hard to find out what’s going on.

It’s as if each subgroup is a different colour: red, green, blue. If they are mixed together, we get a muddy brown, and it’s hard to see the picture.

PrecisionLife’s approach treats subgroups as the solution rather than a problem. It aims to identify groups of patients who share the same disease signature (or overlapping disease signatures). The focus on combinations of SNPs, instead of single ones, and looking for subgroups generates a stronger signal.

Big DNA studies look at how common a single SNP is in patients compared with healthy controls. Typically, there is only a small difference, and having a single SNP might increase the risk of disease by a modest 10-20%.

But the links between combinations of SNPs and the subgroups they define increase the risk of disease by far more – typically fourfold – compared with healthy controls. These stronger associations are easier to find, allowing PrecisionLife to find differences that other genetic approaches can only find with far larger samples.

Dramatic findings awaiting confirmation

Compared with everything published to date, these are spectacular findings. They also come from analysing a very small sample by the standards of genetic research – just 2,400 patients.

Limited success with replication

These are striking results produced by a new method, so it’s natural to be check if the results can be repeated.

The authors tried to do this, using a separate UK Biobank group. This was made up of around 1,300 people who reported in an interview that they had a diagnosis of CFS (rather than being asked if they had ME or CFS).

Success was limited. Five of the 25 critical SNPs were also statistically significant in the replication group, but none of the 84 disease signatures was. The five critical SNPs identified 2 of the 14 genes from the first group.

The paper says that technical reasons meant they were likely to miss at least some of the disease signatures or critical SNPs in the second group of patients. This is something the researchers intend to address in future studies.

The authors also pointed to the slightly different diagnoses for the two groups, which meant that they might not have been comparing like with like.

For now, though, it’s unclear how well these findings replicate.

We should also note that while the researchers describe the method in some detail, PrecisionLife uses a patented process. Effectively, it’s a black box, and other scientists can’t look inside to check its workings. (The authors told me their patented approach reduces the time taken to do these calculations by millions of years!)

Success with other illnesses

But what makes PrecisionLife’s approach so interesting are the results they report for other diseases.

PrecisionLife made the first genetic analysis of Covid, which ran on just 725 patients from the UK Biobank. They found 68 genes of interest and reported that 48 have since been associated with Covid in published papers from other groups.

Examples of the association include:

• Five of their genes of interest were also found by a much larger genetic study of Covid.
• They suggested versions of a gene they identified could play a role in Covid inflammation and a particular drug could reduce inflammation. The gene was shown to be active in the lung cells of severely ill Covid patients and the drug did reduce inflammation.
• The genetic signals they found pointed to the potential of 29 drugs to treat Covid. Thirteen of these have been tested. Most results are not yet available, but one drug has proved effective in a clinical trial.

Issues about the UK Biobank sample

PrecisionLife used data from the UK Biobank as it currently has the largest available of people with ME/CFS. However, it is far from ideal.

The UK Biobank’s diagnosis of ME or CFS might not be very accurate. It is solely based on people saying that a doctor has said that’s what they have, and there are no questions to check that people meet research criteria.

Also, the sample might not be representative. In the first group, the average age was 69, and only 71% were women, compared with 80–85% in most research studies of ME/CFS.

What do the 14 genes do, and can they explain ME?

Back to the study findings. The researchers identified 14 genes associated with ME/CFS. Can these genes – and their biological function – explain why people with particular versions of them are more likely to get ill? Or why people have the symptoms they do?

The authors point out that linking genes to biological function is somewhat open to interpretation. But there is evidence linking the genes to ME/CFS:

1. Autoimmunity

Seven of the 14 genes are linked to autoimmune diseases (there is a lot of genetic overlap between autoimmune diseases), particularly to multiple sclerosis. There is already some published evidence of autoimmune problems in at least some people with ME.

2. Energy metabolism

Several of the genes affect energy metabolism, including one that affects the power stations of the cell, mitochondria, including mitochondria activity after exercise. The effect happens via AMPK, an important molecule that acts as the master regulator of cells’ energy balance. Professor Julia Newton’s team has linked AMPK to ME/CFS.

3. Sleep

Two of the genes affect sleep through our 24-hour circadian rhythms. Poor sleep is a core symptom of ME.

4. Infection

Five of the genes are linked to viral and bacterial infection. Around 70% of people with ME report that their illness began with an infection.

Next moves

Professor Chris Ponting told me he was very interested in these findings. He also said the ME/CFS gene variants identified in this study would “need further genetic support from more conventional studies such as genome-wide association studies”.

As an alternative, he said, he’d like to see if “PrecisionLife’s method can validate the same combinations of DNA variants in independent cohorts of people with ME”.

Happily, DecodeME is already collecting data for a large DNA study, which might confirm some of the genetic links identified here.

And PrecisionLife is in discussions with DecodeME about its own analysis of DecodeME data. DecodeME can provide larger samples with a more accurate diagnosis of ME/CFS.

How these new findings could change the landscape

If the findings from this new study do pan out, we might see rapid progress in ME research and the development of treatments.

Researchers could use the genes highlighted as clues pointing to what is going wrong in ME. That’s a key step to developing new treatments.

Using SNPs to split patients into subgroups could lead to more focused research and clearer findings.

In the same way, SNPs could be used as biomarkers, helping to diagnose ME and even identify which type of ME someone has.

And PrecisionLife has much expertise in identifying drug targets and evaluating which drug candidates are most likely to succeed.

The next critical step is confirmation of these remarkable findings. It could come as soon as next year, depending on DecodeME data availability. And if it does, it will be very good news for people with ME.

I thank Drs Sayoni Das and Krystyna Taylor of PrecisionLife for talking to me about their study.

PrecisionLife presented the findings of this study at the ME Genetics Research Summit on 14th September 2022. A video of the presentation will be available shortly.

The summit marks the launch of a new Genetics Centre of Excellence. Founded by Action for ME and the MRC Human Genetics Unit at the University of Edinburgh, the aim is to create a network of researchers and increase high-quality genetic research and funding for it.

Image credits. DNA, Canstock Photo. Subgroups, PrecisionLife.