Seven replicated genomic associations of myalgic encephalomyelitis/chronic fatigue syndrome: a biobank study

Problem Statement
Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is a debilitating, female-biased illness lacking diagnostic biomarkers, effective treatments, or a well-understood etiology. Previous genetic research has been hindered by small, heterogeneous cohorts and reliance on single-line diagnostic codes, leading to a lack of reproducible findings. No genomic risk loci for ME/CFS had previously replicated across independent cohorts. Furthermore, the role of socioeconomic status (SES) and gene-by-sex interactions in modifying risk remains contentious due to conflicting reports and methodological limitations in prior studies.

Methodology
The authors conducted a genome-wide association study (GWAS) utilizing data from the UK Biobank (UKB) and the All of Us (AoU) Research Program.

• Phenotyping Strategy: Instead of relying on a single diagnostic code, the study employed a multi-evidence phenotyping strategy to define cases and controls. Cases were required to have self-reported clinical diagnoses, poor/fair overall health ratings, and specific affirmative responses to pain questionnaires regarding ME/CFS. Controls were required to have good/excellent health ratings and no evidence of ME/CFS in medical records or surveys.
• Cohorts: Three disjoint populations were established:
  1. Discovery (UKB1): 1,268 cases and 113,132 controls.
  2. Replication 1 (UKB2): 319 cases and 28,471 controls.
  3. Replication 2 (AoU): 371 cases and 33,140 controls.
• Statistical Framework: The study utilized TarGene (Targeted Genomic Estimation), a semi-parametric, doubly-robust estimation workflow based on Targeted Learning. This approach was chosen to maximize power while minimizing bias from model misspecification and population stratification, addressing limitations of conventional parametric GWAS methods.
• Analysis Pipeline:
  • Discovery: A GWAS was performed on 670,243 genotyped SNPs in UKB1, controlling the False Discovery Rate (FDR) at 5%.
  • Replication: The 176 FDR-significant variants from the discovery phase were tested in the two replication cohorts.
  • Fine-mapping and Colocalisation: For replicated loci, fine-mapping was performed using SuSiE to generate credible sets of causal variants. These were colocalised with expression quantitative trait loci (eQTLs) from the GTEx v10 database.
  • Interaction Analysis: Gene-by-sex and gene-by-deprivation (using Townsend Deprivation Index and Indices of Multiple Deprivation) interactions were tested using TarGene.
  • Validation: A look-up was performed in the DecodeME cohort (2,143 cases), a study using strict clinical criteria (Canadian Consensus and IoM 2015) which requires Post-Exertional Malaise (PEM).

Key Results

• Discovery: The analysis identified 176 variants significantly associated with ME/CFS risk (FDR < 5%) in the UKB1 discovery cohort.
• Replication: Seven genomic risk loci replicated in at least one of the two disjoint replication cohorts (UKB2 or AoU). However, no single variant achieved significance across all three cohorts after multiple testing correction.
  • The seven replicated variants are: rs115186419 (near ETV5), rs73175505 (near CSMD1), rs261902 (in BICD1), rs117553493 (near CLYBL), rs72741654 (near RORA), rs74963073 (in GRIN2A), and rs76847656 (near SBK1).
• Fine-mapping and Colocalisation:
  • At the chromosome 13 locus, fine-mapping resolved a credible set colocalising with reduced CLYBL expression in the putamen (basal ganglia). The risk allele is associated with lower CLYBL expression (Posterior Probability of H4 = 0.994).
  • Notably, the CLYBL Arg259 stop-gain variant (a known loss-of-function variant) was not associated with ME/CFS risk, suggesting the signal is driven by regulatory variants affecting expression dosage rather than protein truncation.
• Interactions: No gene-by-sex or gene-by-deprivation interactions survived multiple-testing correction. However, nominal signals were observed for sex-differential effects at rs261902 and rs76847656, and a deprivation-differential effect at rs73175505.
• DecodeME Validation: None of the four testable replicated variants showed association in the DecodeME cohort. The authors attribute this discrepancy to differences in phenotype definition (DecodeME requires PEM; this study did not), age distributions, and recruitment strategies (community vs. biobank).

Significance and Claims
The paper claims to be the first ME/CFS study to report genetic associations that replicate in independent cohorts, identifying seven candidate loci. The authors emphasize that the use of multi-evidence phenotyping and the TarGene statistical framework allowed for the detection of signals that previous single-code studies missed.

The study modestly concludes that while no variant replicated across all three cohorts (likely due to phenotypic heterogeneity, population differences, and the "winner's curse" in smaller replication cohorts), the seven replicated loci provide candidate targets for future biological investigation. The colocalisation of the chromosome 13 signal with CLYBL expression in the putamen highlights a potential link between mitochondrial vitamin B12 metabolism and ME/CFS pathophysiology. The authors explicitly state that these findings do not constitute causal proof but offer a foundation for follow-up studies into the biological mechanisms of ME/CFS. They caution that the lack of overlap with DecodeME suggests that genetic architecture may differ depending on how the disease is defined (e.g., inclusion of PEM).