Precision medicine biomarkers for Myalgic Encephalomyelitis / Chronic Fatigue Syndrome (ME/CFS) using multi-modal AI/ML

About the Project

Title: Precision medicine biomarkers for Myalgic Encephalomyelitis / Chronic Fatigue Syndrome (ME/CFS) using multi-modal AI/ML

Synopsis: Can we leverage precision medicine to identify diagnostic markers that correlate with ME/CFS symptom severity or disease activity? Are ME/CFS blood-based diagnostic biomarkers being missed because each person has their own set of haematological setpoints?

Details: Identifying biomarkers of disease activity for Myalgic Encephalomyelitis/Chronic Fatigue Syndrome would represent a step-change in our ability to diagnose cases, understand the pathophysiology and provide objective outcome measures in future clinical trials. Recently, we discovered a set of 116 molecular or cellular traits that significantly differ, on average, between people with ME/CFS and others, first for females and then separately for males (Beentjes et al. 2025, above). This project seeks to build on these discoveries by investigating whether people with ME/CFS show plasma protein differences that correlate with their self-reported symptom severity.

This project was also motivated by findings that an individual’s haematological measurements fluctuate little around a stable value (their setpoint) and that measurements vary more substantially from patient to patient (https://doi.org/10.1038/s41586-024-08264-5). Because they depend on cell type abundance, we hypothesise that plasma protein abundances similarly have setpoints. Consequently, departure from a setpoint due to disease might be diagnostic, but this would only become evident upon analysing multiple data points from the same individual, longitudinally.

Deviations from setpoints may be particularly important in ME/CFS since the cardinal symptom is post-exertional malaise (PEM): the exacerbation of symptoms or onset of new symptoms following exertion. The fluctuating and heterogeneous symptom profiles pose a challenge to studying ME/CFS since changes in biomarkers may only be detectable transiently. This project aims to leverage at-home longitudinal sampling and state-of-the-art statistical methods to identify biomarkers that correlate with fluctuating symptoms and disease activity.

We will leverage a cutting-edge method, Longitudinal Targeted Minimum Loss-based Estimation (LTMLE), to correlate protein expression with symptom severity over time, whilst accounting for time-varying confounders (Van der Laan & Gruber, LTMLE R package, Shirakawa et al.). In particular, we will define whether protein measurements fluctuate significantly with symptom severity and extract a minimal and predictive subset across all participants.

Participants will be recruited via DecodeME: 95% of the 21,620 DecodeME participants consented to being recontacted for new research projects. We will recruit hundreds of DecodeME participants with an ME/CFS diagnosis. Samples will be collected as dried blood spots, provided in participants’ homes, and protein levels would be measured via Olink proximity extension assays. This approach is both feasible (Fredolini et al.) and desirable given that 25% or more of people with ME/CFS are housebound or completely bedbound (Prendergrast et al.). We will initially measure protein levels from 34 selected individuals who have provided blood samples for 5 timepoints taken on days with either high or low symptom burden with matching questionnaire data. Any additional dried blood spot samples will be stored for our and others’ future research.

We are particularly interested in focusing on the 16.3% of people who develop ME/CFS after Glandular Fever. This is because – like in Long Covid – they form a homogeneous cohort, all of whom experienced an Epstein-Barr Virus (EBV) or Cytomegalovirus (CMV) infection prior to onset of their chronic symptoms. However, results could potentially be relevant to other viral and non-viral triggers.

The DecodeME study placed people with lived experience at the heart of its science, and this project will do the same. It will use the patient and public involvement pool that will be set up by the PRIME project (lead: Ponting) by May 2026.

Training: Methodologically, the student will develop technical skills in the development and application of rigorous statistical inference (semi-parametric efficiency theory) and machine learning techniques throughout the PhD and by auditing MSc level courses in these areas and beyond. In the application of biomedical data at various scales, on the biomedical front, the student will develop an understanding of high-dimensional and temporal molecular biology via proteomics and disease phenotypes. The student will further develop essential cross-disciplinary and translational communication with access to a supervisory team with diverse expertise ranging across AI/ML, biostatistics and molecular biomedicine.

The student will receive training in patient and public involvement (PPI) and in communicating their research to lay audiences. Optionally, they would have the opportunity to develop surveys for data collection and could gain experience with wet lab work although this is not required. We could support them with open science research practices such as study pre-registration on the Open Science Foundation.

Recruitment: Experience with programming in R and a basic knowledge of statistics, as well as an interest to expand these skills, is required.

Apply: All applications must be submitted through the Future Medicine PhD fellowships website.

Funding Notes

Students will receive a stipend at UKRI levels, plus £30K in travel and research funds across all three years of the fellowship. All University fees will be covered.

The fellowships are open to students who are eligible for home fees at Edinburgh - i.e. you must be a UK national, or have settled status, and have. been "ordinarily resident" in the UK for the three years immediately before the start of the fellowship. Other international applicants are not eligible for these fellowships.