对系谱、流行病学和分子数据的多维度分析为肌痛性脑脊髓炎/慢性疲劳综合征提供病因学线索

Background: Myalgic encephalomyelitis (ME)/chronic fatigue syndrome (CFS) is a complex disabling disorder with no known etiology or approved treatment. Estimates of the prevalence suggest that up to 3.4 million Americans may be afflicted and emerging evidence indicates that the COVID-19 pandemic may lead to a significant increase in ME/CFS cases globally. We conducted a molecular epidemiologic study to identify risk factors and biologic mechanisms for ME/CFS. Methods: Our clinic-based case-control study involved 60 carefully selected ME/CFS patients and 61 appropriately matched healthy controls. We compared cases and controls with respect to the following: 1. prevalence of autoimmune disease (AID) and cancer among their first-degree relatives, 2. prevalence of epidemiologic factors, 3. serum levels of 48 cytokines, and 4. whole-blood RNA-seq gene expression data. We used conventional and machine learning approaches to calculate associative and predictive metrics, and to identify cytokine and gene expression profiles of ME/CFS. Results: First-degree relatives of ME/CFS cases were more likely than those of the controls to have AID [Relative Risk (RR)=3.52, p=0.0014) and early-onset (diagnosed <60 years of age) cancer (RR=2.24, p=0.034) including blood cancers (p=0.047). Comparison of epidemiologic factors identified several risk factors such as history of allergies requiring medication [Odds Ratio (OR)=6.00, p<0.0001), exposure to contaminants (OR=4.35, p=0.0002), history of illness requiring hospitalization (OR=4.33, p=0.0004), ≥4 episodes of significant illness requiring hospitalization (OR=24.36, p<0.0001), and ≥2 episodes of significant stress (OR=3.07, p=0.03). The most common self-identified perceived causes of ME/CFS reported by cases in response to an open-ended question were Infectious Illness (27.3%), Infectious Agents (15.9%), and Stress (15.9%). We identified a cytokine signature of ME/CFS, which classified patients with AUC>0.75, sensitivity>80%, and specificity>70% at an optimized threshold in all three tested machine learning models: XGBoost, k-nearest neighbors, and Support Vector Machines. Key cytokine predictors included IL-27, IP-10, RANTES, and Fractalkine among others. Whole blood RNA-seq analysis identified 115 differentially expressed genes with FDR<0.25 belonging to biologic pathways relevant to infectious diseases and neurologic disorders. Conclusions: Our multidimensional analysis identified previously unreported risk factors for ME/CFS, links with AID and early-onset cancer, a dysregulated immune profile, and potential biologic mechanisms-such as neuronal injury-thus providing etiologic clues and druggable targets for treatment.