The use of artificial intelligence (AI) has rapidly advanced in recent years, enhancing our ability to process large amounts of data, identify patterns, and make useful predictions. AI takes advantage of large amounts of data and can rapidly search for patterns, utilise the data for specific outputs, and predict future possibilities based on information available and patterns noticed. AI is being harnessed in ME/CFS research, as we have previously reported on, in the search for biomarkers and evidence for the biological basis of the disease. A recent study introduced a new AI tool, BioMapAI, developed collaboratively by scientists at Duke University School of Medicine and the Jackson Laboratory. The tool was trained on four years of biological and clinical data from 249 participants and, according to a Duke University School of Medicine article, can identify ME/CFS with 90% accuracy by analysing stool, blood, and routine lab tests.
Dr Julia Oh, microbiologist and senior study author, explained “We integrated clinical symptoms with cutting-edge -omics technologies to identify new biomarkers of ME/CFS”. Also on the team were Dr Lucinda Bateman and Dr Suzanne Vernon of the Bateman Horne Center. The study found “patients with ME/CFS had lower levels of butyrate, a beneficial fatty acid produced in the gut, and higher levels of tryptophan and benzoate, markers of microbial imbalance… Their immune systems also showed heightened inflammation, particularly in MAIT cells, which link gut health to immune function.” Whilst further research is needed to confirm the results, scientists note that this study represents a major step forward in understanding ME/CFS and provides direction for future investigations.
The study in more detail
Aim
The primary goal of the study was to identify disease biomarkers for ME/CFS, including those specifically associated with its heterogeneous (diverse) symptoms, and to map interactions among the gut microbiome, immune system, and metabolome (complete set of small molecules within a biological sample). This multi-omics approach considering multiple data types together moves beyond single or pairwise data analyses, which look at one or two data types at a time respectively.
BioMapAI Explained
BioMapAI is described as a supervised deep neural network trained on multi-omics data. Breaking these terms down:
- Neural network: A computer system, inspired by human brain networks, that can detect complex patterns in data.
- Supervised: A model trained by being shown the “right answers”, such as which participants have ME/CFS and which are healthy, allowing it to learn patterns which aid predictions.
- Multi-omics: Integration of multiple types of biological data, including microbiome composition, blood metabolites, immune cell profiles, and lab results.
Together, this means BioMapAI learns from different types of biological data over time to identify patterns linked to ME/CFS.
Methods
The study generated and assembled a dataset from 249 participants, whom they followed over 3-4 years, including 153 participants with ME/CFS (diagnosed according to the IoM criteria) with a disease duration of <4years and 96 age- and gender- matched healthy controls, collecting blood and stool samples and detailed clinical data. Eventually the dataset consisted of details about the gut microbiome, blood metabolites, immune cells, and lab data. The researchers constructed a unique connectivity map from data “adjusted for age, gender and additional clinical factors” which uncovered associations. Key findings were also validated in external cohorts to strengthen the results.
BioMapAI Findings
The study found that people with ME/CFS have changes in their gut bacteria, metabolism, and immune system that correlate with their symptoms. Beneficial gut microbes, such as Faecalibacterium prausnitzii, were reduced, along with alterations related to important molecules microbes produce, such as butyrate, benzoate, tryptophan and branched-chain amino acids. At the same time, there were heightened proinflammatory responses, meaning the immune system is more active than normal and producing extra signals that promote, or are associated with, inflammation. These biological changes were associated with fatigue, sleep problems, emotional dysregulation, and reduced social activity. Overall, whilst the findings highlight correlations rather than causality, they suggest that ME/CFS symptoms arise from disrupted connections between the gut, metabolism, and immune system.
Limitations
According to the researchers, “This dataset is among the most comprehensive multi-omics resources assembled for ME/CFS (including other complex chronic diseases)” nevertheless, they acknowledge that the sample size was still relatively small given the complexity of training an AI model. They suggest that the training dataset needs to be expanded and further validation done.
They also acknowledge, “our study population comprised more females and older individuals, most of whom were white (though this is consistent with the epidemiology of ME/CFS10), and was from a single geographic location (Bateman Horne Center), which may limit our findings to certain populations.”
Conclusion
These preliminary findings build on the growing knowledge we have about ME/CFS and other work being done with AI which may help pin down key disease patterns and aid the development of targeted treatments for ME/CFS. Further research needs to be done to validate the study findings.
