Reduced FADS1 Activity and How Severe MS Becomes
Most genetic work in multiple sclerosis (MS) asks who develops the disease. Noroozi and colleagues asked a harder question: which blood metabolites causally affect how severe it becomes. They built the analysis in five layers. A metabolome-wide Mendelian randomization scan of 1,091 metabolites and 309 ratios, instrumented from a GWAS of 8,299 participants, against the age-related MS severity score in 12,584 people with MS from the International MS Genetics Consortium, a cohort enriched for older patients with longer disease duration at a mean of 18.2 years. Then multivariable MR within metabolite classes, genetic colocalization at the loci that came up, cis-MR using functional variants that proxy specific gene perturbation, and finally expression-based MR across seven brain cell types. This is a medRxiv preprint and has not been through peer review, which is worth holding in mind throughout.
The Primary Screen Is Weaker Than the Headline Suggests
The methods pre-specify a significance threshold of P < 6.8×10⁻⁴, derived from an effective number of 73 independent metabolites. The results then report 45 metabolites with nominally significant estimates at P < 0.05, of which 36 were annotated and nine were unidentified compounds. Checking those estimates against the study's own threshold, exactly one clears it: cis-3,4-methyleneheptanoylglycine at P = 3.58×10⁻⁴. The remaining 44 sit between roughly 0.003 and 0.048. The paper is open about this, labelling them nominally significant rather than significant, but it does mean the entire downstream architecture rests on a screen that did not clear its own bar. Two further choices compound the problem. Instruments were selected at P < 1×10⁻⁵ rather than genome-wide significance, which the authors justify by the modest size of the metabolite GWAS, and that GWAS has 8,299 participants, which is small for the purpose.
The Pathway Analysis Is Where the Signal Actually Is
Multivariable MR within metabolite classes produces the study's strongest numbers, and here two results clear the pre-specified threshold with room to spare. Arachidonate (20:4n6) came in at β = −0.07 with P = 7.78×10⁻⁷ and betaine at β = −0.07 with P = 1.56×10⁻⁵, both associated with lower severity. Arginine at β = 0.07 and P = 9.13×10⁻⁴ was associated with higher severity, landing just outside the threshold, with propionylglycine and succinoyltaurine weaker at 0.04 and 0.015. What matters about this step is the direction of travel: accounting for correlation between related metabolites sharpened the lipid signal rather than dissolving it. When a screen result is an artifact of correlated instruments, adjustment usually does the opposite.
The FADS Locus, and What Colocalization Cannot Resolve
Colocalization on chromosome 11 returned posterior probabilities above 0.70 for a shared causal variant across three fatty acid ratios, all of them involving arachidonate. The problem is one the authors raise themselves: this is a well-established pleiotropic locus holding multiple overlapping genes, and the signal colocalizes with FADS1, FADS2, MYRF, and TMEM258 alike. The putative shared variant for FADS1/2 is rs174564, while rs174537 and rs102274 came up for MYRF and TMEM258, and those two sit in near-complete linkage disequilibrium with rs174564 at D′ = 1.0 and R² = 0.99. Statistically they cannot be told apart. The assignment to FADS1 therefore rests on biological plausibility and on expression data rather than on the colocalization result, and the paper says as much.
The Expression Argument That Makes the Gene Assignment Stick
The case for FADS1 over its neighbours comes from a functional variant rather than a statistic. rs174546 C>T sits in the 3′ untranslated region of FADS1, and the T allele creates a miR-149-5p binding site, producing allele-specific repression of the gene. Expression data show the TT genotype associated with reduced FADS1 across multiple brain regions, liver, and whole blood, while FADS2, MYRF, and TMEM258 show no significant reduction and are generally modestly increased. Using that variant as a single instrument, the metabolite pattern comes out coherent across roughly 25 separate measurements: reduced Δ5-desaturase activity, upstream FADS2-derived intermediates accumulating, downstream FADS1-derived products falling, and greater MS severity at every step. Directional consistency across an entire biosynthetic cascade is a different and stronger kind of evidence than any single estimate in the original screen.
Down to Individual Brain Cell Types
Using the same variant as an expression quantitative trait locus across seven brain cell types in 474 individuals, reduced FADS1 expression was associated with greater MS severity in astrocytes (β = 0.153, P = 1.3×10⁻²), oligodendrocytes (0.112, 1.2×10⁻²), excitatory neurons (0.101, 9.4×10⁻³), and inhibitory neurons (0.0766, 7.8×10⁻³). No association appeared in endothelial cells, microglia, or oligodendrocyte precursor cells. The selectivity is informative on its own terms, since oligodendrocytes and neurons are precisely the cells whose membranes depend most on long-chain polyunsaturated fatty acids. These p-values are nominal and uncorrected across the seven cell types tested, so the pattern of which cells respond carries more weight here than any individual estimate does.
The Second Locus, and What Would Test All of This
A separate signal appeared at CYP4F2, where the shared variant is rs2108622, a missense change (p.Val433Met) that lowers CYP4F2 protein without altering transcription or enzymatic activity, and where genetically proxied reductions in activity tracked with greater severity. The colocalization is weaker here, at a posterior probability above 0.4 rather than 0.7, which the authors describe as moderate. Their limitations section is thin for a paper of this complexity, naming European-only ancestry and the absence of sex- and age-stratified analyses. Two constraints belong alongside those. The outcome is a severity score in 12,584 patients, and severity GWAS remain far less powered than susceptibility ones. And because the primary screen leaned on nominal significance, the burden of proof falls almost entirely on the convergence of colocalization, cis-MR, and cell-type expression pointing the same way. That convergence is the paper's actual argument, and it can be tested directly: measure the fatty acid cascade in patients stratified by rs174546 genotype and follow their severity over time, which would show whether this holds in people rather than in summary statistics.
Disclaimer: This blog post is based on the cited preprint and is intended for informational purposes only. The preprint has not been certified by peer review. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.
Reference:
Noroozi, R., Higgins Tejera, C., Chen, M., Briggs, F. B. S., Bhargava, P., & Fitzgerald, K. C. (2026). Integrative genetic analyses of lipid metabolism and multiple sclerosis severity using metabolome-wide and cis-Mendelian randomization. medRxiv. https://doi.org/10.64898/2026.05.27.26354239
