NEW YORK – A proprietary hypothesis-free analytics method from PrecisionLife has uncovered a set of genes potentially related to risk of developing myalgic encephalomyelitis, also known as chronic fatigue syndrome. The UK firm hopes the signatures can now be used to repurpose existing medicines and to develop animal models of ME/CFS.
Steve Gardner, CEO of PrecisionLife, said that the firm's combinatorial analytics is a novel mathematics-based method that analyzes the effects of interactions between many genetic mutations.
"We look for a different set of signals than you can identify using genome-wide association studies," he said in a recent interview. Where GWAS queries relatedness between one particular SNP and a patient population, the combinatorial method detects the combined interactions of many. The computations take between four hours to a week to complete but only require fairly standard computing power, he said.
The UK company has targeted so-called complex chronic disorders with its method, such as ME/CFS, Alzheimer's disease, and sepsis, in which multiple genetic variations might come together to exert a particular effect on metabolism.
For ME/CFS in particular, the genetic underpinnings have been elusive to date. There are no animal models of the illness and no laboratory tests to conclusively diagnose it.
Patients can spend years getting a diagnosis, Gardner said, and the underlying stigma is "a massive burden" to people who are already suffering with debilitating symptoms. Worldwide, there are an estimated 20 million sufferers, "with a 100 percent unmet medical need," Gardner also said.
PrecisionLife's team applied its proprietary combinatorial analytics method to a genetic dataset of 2,382 ME/CFS patients from the UK Biobank and 4,764 controls.
The team uncovered 199 SNPs mapping to 14 genes, and these could be clustered into four pathways. Interestingly, a single SNP analysis of this same population revealed no significant genes.
Regarding the pathways, "they are all highly plausible in terms of the pathophysiology of ME/CFS," Gardner said.
The patient stratification engendered by these four clusters is also novel, although clinicians have anecdotally reported similar clinical phenomena, he said.
The four categories included genes related to the immune response to stress and infection, energy metabolism and mitochondrial dysfunction, sleep disturbances and circadian rhythm dysfunctions, and genes associated with autoimmunity.
The research was published in medRxiv earlier this month along with a high-level description of the combinatorial analytics method. However, "the details of the computational methods that make it possible are trade secret," Gardner said.
The newly discovered risk genes may now help narrow down disease-associated pathways for further investigations and to better select patients for clinical trials.
Next steps
ME/CFS is currently hypothesized to be triggered by an infection. Previous metabolomics research has suggested that the illness could be akin to a post-infection or post-stress metabolic state that failed to return to normal. Indeed, recent research has also discovered that some patients with severe ME/CFS show elevated serum immunoglobulin A antibodies to a particular protein in the flagellum of an abundant gut microbe.
Although the risk genes PrecisionLife uncovered aren't diagnostic, they may be useful to "home in on the pieces of metabolism that may give a signal we can detect in the blood or through some noninvasive measure," Gardner said, to potentially give a firm diagnosis.
Still, the lack of diagnostic tests has meant that it is hard to assemble an ME/CFS patient population for medication trials.
PrecisionLife plans to use the risk genes for a "precision repurposing analysis" of existing pharmaceuticals targeting the implicated genes and pathways, to identify any that might be helpful to ME/CFS patients in the short term, Gardner said.
Novel drug targets can also be pursued with these genes, should they prove to be "druggable," he said, but in this case the lack of animal or cellular models for ME/CFS makes it a more challenging project.
"We don't even know what cells are involved in ME/CFS, so setting up assays systems is really in its infancy," he said.
Thus, another immediate benefit of the study might be to direct research toward genes that can be manipulated, and the particular cell types in which those genes are expressed, to recapitulate the disease.
A similar case can be made for the disease now called long COVID, and Gardner noted that his team has found overlap between the genes associated with ME/CFS, long COVID, fibromyalgia, post-viral syndrome, and multiple sclerosis. "It may be that medications we can identify as effective in one of these, may be effective across several of them," he said.
Chris Ponting, a genetic researcher at the University of Edinburgh and principal investigator of the DecodeME study, reviewed early versions of PrecisionLife's manuscript.
In an email, he said, "As it is an unconventional analysis, it's difficult to fully assess its results or to estimate how many of its predictions are likely to be replicated in the future."
Nevertheless, should the same DNA combinations be highlighted in independent analyses, it might be more convincing, he said. If this occurred in multiple independent analyses "then this would be a game-changer for ME/CFS," Ponting said.
"Knowing what system is going wrong in people with ME/CFS — whether it is their immune or nervous system, for example — would allow more targeted research to be prioritized and other unsubstantiated lines of enquiry to be curtailed," Ponting also said.
Unfortunately, he added, to date no ME/CFS experimental result has been replicated, noting that "replication is key: without it a reliable diagnostic test will not be developed."
Ron Davis, a genomics pioneer and the director of the ME/CFS Collaborative Research Center at Stanford University, said in an interview that the PrecisionLife study is "very interesting."
Where others have looked at the same biobank data without finding very much, the PrecisionLife team "found a lot," Davis remarked.
Davis' team has developed a yeast model of a potential ME/CFS pathological mechanism involving tryptophan pathways, as well as a diagnostic device that measures ATP-deficits in response to stress in a patient's white blood cells.
He concurred with Ponting, however, that replication of the PrecisionLife ME/CFS results will be key.
Ideally, the mathematical underpinnings of the method should be scrutinized by experts, Davis said, but with a proprietary computational method the next best option may be confirmation of the results by other labs. Indeed, some of the pathways PrecisionLife claims to have detected jibe with his lab's results, Davis said, adding that the links to fibromyalgia and multiple sclerosis are particularly intriguing.
The next step in the ME/CFS work will be to apply the method to other biobanks, including DecodeME, Gardner said.
Ponting noted that the DecodeME trial is currently recruiting up to 25,000 people with a diagnosis of ME/CFS who live in the UK. "The large size of the DecodeME study means that many initial predictions, including those made by PrecisionLife, can rapidly be tested for replication."
PrecisionLife has already run the analysis on sample sets representing more than 40 other complex chronic diseases, "everything from Alzheimer's and ALS, through coronary artery disease, endometriosis, and rheumatoid arthritis," Gardner said.
In May 2020, the firm ran its combinatorial analytics method on samples from SARS-CoV-2-infected patients to examine the host response genetic signature of SARS-CoV-2 infection. Of the 68 genes implicated, 48 have been since reported in the literature, Gardner said.
The firm also discovered six subgroups in Alzheimer's disease with different mechanisms of action, and found that approximately 70 percent of patients had attributes of more than one mechanism.
The firm is now performing two new studies each month, and has so far discovered more than 300 novel targets and indication extension opportunities using patient stratification biomarkers.
It has also formed numerous target discovery and validation R&D collaborations with biotech and pharmaceutical firms, Gardner said. As part of this, it has a workflow for "systematic precision repositioning" whereby it goes through the pipelines of major pharma companies to identify other indications for their compounds.
"We also do a lot of analysis around clinical trials, either designing patient stratification biomarkers and exclusion criteria, so you can design more targeted trials, or even analyzing Phase III clinical trial data to evaluate biomarkers of drug response," Gardner said. It could be that many drugs that "failed" clinical trials worked for subgroups, he said, and that clinical efficacy could now be demonstrated with targeted trials.
Stepping back, Gardner said that "anytime you are dealing with such huge amounts of data, at a minimum you have to have incredibly good statistics married to a really good understanding of how human genetics and biology work."
If a gene is only relevant in a quarter of patients, "you are not going to find it with GWAS," he said. But, "we can."