Investigation of Genetic Variants and Causal Biomarkers Associated with Brain Aging

Jangho Kim, Junhyeong Lee, Seunggeun Lee

TL;DR

Researchers used a multimodal approach—combining 3D deep learning, genomics, and causal inference—to investigate "delta age," the discrepancy between a person's chronological age and their brain's biological age. The study identifies the fornix and thalamus as critical regions for predicting brain aging and provides evidence that immune-related blood biomarkers aren't just correlated with aging, but may actually cause it.

What Problem Does This Paper Address?

We all know people who seem "young for their age," and the same applies to our brains. "Delta age" is a metric used to quantify this: if a 60-year-old has the brain structure of a 70-year-old, they have a high delta age, which is often a precursor to neurodegenerative diseases like Alzheimer’s.

While AI can now predict brain age from MRI scans with high accuracy, the "why" remains elusive. We don't fully understand which specific brain regions drive these predictions, which genes are responsible, or whether systemic factors (like blood chemistry) are causes or merely symptoms of an aging brain. This paper seeks to move from "black box" prediction to an explainable, causal model of brain aging.

Methods

The authors employed a sophisticated three-tier pipeline:

  1. Explainable AI (XAI): They trained a 3D Convolutional Neural Network (CNN) on T1-weighted structural MRIs to predict age. To understand the model's "logic," they used Integrated Gradients (IG), a saliency mapping technique that highlights which specific voxels (3D pixels) in the brain image most influenced the age prediction.
  2. Advanced Genomics: They performed Genome-Wide Association Studies (GWAS) using both standard SNP arrays and Whole-Exome Sequencing (WES). While SNPs look at common variations, WES allows for a "gene-based" test (using SKAT-CommonRare) to identify rare variants that might have a large impact on brain aging.
  3. Causal Inference: To move beyond simple correlation, they used Mendelian Randomization (MR). By using genetic variants as "natural experiments," they tested whether 310 different blood and metabolomic biomarkers (like cholesterol or white blood cell counts) have a direct causal effect on increasing delta age.

Datasets

The study relies on the UK Biobank, one of the world's most comprehensive biomedical databases.

  • MRI Data: 34,129 samples (White British) were used for the analysis.
  • Genomics: 38 million imputed genetic variants and 200k Whole-Exome Sequencing records.
  • Biomarkers: 249 metabolomic phenotypes and 61 blood-related phenotypes.
  • Training Set: 25,656 healthy individuals were used to train the initial age-prediction model.

Key Insights and Ideas

  • The Fornix and Thalamus are "Canaries": While previous studies pointed to the hippocampus, this study’s XAI approach found that lower volumes in the fornix and the lower part of the thalamus are the strongest predictors of an "older" brain. These regions are vital for memory and sensory processing.
  • Shared Genetic Architecture: The researchers found a high genetic correlation between delta age and the volume of these specific brain regions, suggesting that the same genetic "program" driving brain shrinkage is also driving the AI's age prediction.
  • The Immune-Brain Connection: Perhaps the most striking insight is the causal link between the immune system and brain aging. The study suggests that systemic inflammation—specifically related to white blood cells—is a driver of accelerated brain aging, not just a byproduct of it.
  • Novel "Aging" Genes: The study identified genes like OMA1 (linked to mitochondrial health and neurodegeneration), NOSTRIN (nitric oxide trafficking), and CPNE3 (carcinogenesis) as significantly associated with delta age.

Key Results

  • Prediction Accuracy: The 3D CNN achieved a Mean Absolute Error (MAE) of 2.64 years in healthy individuals, which is on par with state-of-the-art models.
  • Heritability: The study calculated that delta age has a genetic heritability of 21.5%.
  • Causal Biomarkers: Four immune-related biomarkers passed rigorous multiple-testing corrections in the Mendelian Randomization analysis: Eosinophil count, Eosinophil percentage, Neutrophil count, and Total protein. High levels of these markers were shown to causally increase delta age.
  • Genomic Discoveries: The gene-based test on WES data identified 21 significant genes, including several (like OMA1 and NOSTRIN) never before reported in brain delta age GWAS.

Limitations and Future Work

The authors acknowledge two primary limitations:

  1. Ancestry Bias: The analysis was limited to individuals of European ancestry. Because genetic variants and aging processes can vary across populations, these results may not be globally generalizable.
  2. Lack of External Validation: While the UK Biobank is massive, the study lacks a completely independent external dataset to replicate the findings, primarily because few other datasets offer the same level of multimodal (MRI + WES + Metabolomics) data.

Future Directions: The researchers suggest that future work should investigate the specific mediating factors that allow immune responses to trigger brain atrophy, potentially opening doors for anti-aging interventions targeting the immune system.

Summary updated

Abstract

Abstract Delta age is a biomarker of brain aging that captures differences between the chronological age and the predicted biological brain age. Using multimodal data of brain MRI, genomics, and blood-based biomarkers and metabolomics in UK Biobank, this study investigates an explainable and causal basis of high delta age. A visual saliency map of brain regions showed that lower volumes in the fornix and the lower part of the thalamus are key predictors of high delta age. Genome-wide association analysis of the delta age using the SNP array data identified associated variants in gene regions such as KLF3-AS1 and STX1. Mendelian randomization (MR) for all metabolomic biomarkers and blood-related phenotypes showed that immune-related phenotypes have a causal impact on increasing delta age. Our analysis revealed regions in the brain that are susceptible to the aging process and provided evidence of the causal and genetic connections between immune responses and brain aging.
BibTeX
@inproceedings{kim2023investigation,
  author = {Jangho Kim and Junhyeong Lee and Seunggeun Lee},
  title = {{Investigation of Genetic Variants and Causal Biomarkers Associated with Brain Aging}},
  booktitle = {Scientific},
  year = {2023},
  doi = {10.1038/s41598-023-27903-x},
}
Figure