# Mendelian Randomization and Double Machine Learning Modeling Reveal Brain Imaging‐Derived Phenotypes as Functional Contributors to 18 Autoimmune Inflammatory Diseases

## TL;DR

This study finds that brain activity patterns may influence the risk of autoimmune diseases like multiple sclerosis and lupus, suggesting potential new treatment targets.

## Contribution

The study introduces a novel integration of Mendelian randomization and double machine learning to identify causal links between brain imaging phenotypes and autoimmune diseases.

## Key findings

- Reduced left striatal activity is linked to increased multiple sclerosis risk (OR = 0.59).
- Left uncinate fasciculus activity elevates systemic lupus erythematosus risk (OR = 3.72).
- Asymmetric cerebellar peduncle effects are observed in cutaneous vasculitis (left: OR = 0.11; right: OR = 8.57).

## Abstract

Autoimmune inflammatory diseases (AIDs) are genetically linked disorders with unclear causal links to brain functional networks. Using bidirectional two‐sample Mendelian randomization (MR) on GWAS data from 18 AIDs and 1,366 brain imaging‐derived phenotypes (n = 8,428), we identified significant associations, including reduced left striatal activity increasing multiple sclerosis risk (OR = 0.59), left uncinate fasciculus activity elevating systemic lupus erythematosus risk (OR = 3.72), and asymmetric cerebellar peduncle effects in cutaneous vasculitis (left: OR = 0.11; right: OR = 8.57) [exploratory finding with 24.8%–37.8% power]. Fibromyalgia suppressed cerebellar area VIIIa (β = −0.023). Sensitivity analyses, double machine learning, and >99% statistical power supported robustness. These findings suggest alterations in default mode, salience, and central executive networks contribute to AIDs pathogenesis, highlighting brain regions such as the striatum and cerebellar peduncles as potential therapeutic targets.

This schematic integrates the eight statistically significant causal relationships identified between 1,366 brain imaging‐derived phenotypes (IDPs) and 18 autoimmune inflammatory diseases (AIDs). Arrows indicate the direction of causality inferred from bidirectional two‐sample MR analyses. Statistical tests: Causal estimates are derived using inverse variance weighted (IVW) MR as the primary method, with sensitivity analyses including MR‐Egger, weighted median, MR‐PRESSO, and leave‐one‐out tests. DML validation employed orthogonalized cross‐fitting with Lasso and Random Forest regressors. Data presentation: Results are shown as ORs or β values with 95% confidence intervals. Significance thresholds are Bonferroni‐corrected (forward MR: P < 5.75 × 10−4; reverse MR: P < 1.67 × 10−2). No significance symbols are used in this figure.