ENIGMA and the individual: Predicting factors that affect the brain in 35 countries worldwide

Jieping Ye, Ieee Fellow & Acm Distinguished Member, Paul M. Thompson, Ole A. Andreassen, Alejandro Arias-Vasquez, Carrie E. Bearden, Premika S. Boedhoe, Rachel M. Brouwer, Randy L. Buckner, Jan K. Buitelaar, Kazima B. Bulayeva, Dara M. Cannon, Ronald A. Cohen, Patricia J. Conrod, Anders M. Dale, Ian J. Deary, Emily L. Dennis, Marcel A. de Reus, Sylvane Desrivières, Danai Dima, Gary Donohoe, Simon E. Fisher, Jean-Paul Fouche, et al.NeuroImage 2017 · DOI

TL;DR

The ENIGMA consortium is a massive global collaboration analyzing brain imaging and genetic data from tens of thousands of individuals to uncover factors influencing brain structure and function. This paper reviews their progress in identifying genetic and disease-related influences, highlighting their success in finding subtle but consistent effects worldwide and their potential for future personalized medicine.

What Problem Does This Paper Address?

Understanding the complex interplay between genetics, brain structure, and disease is a major challenge in neuroscience and medicine. Individual brain scans and genetic data hold immense potential for personalized healthcare, but identifying the subtle factors that influence brain health and disease risk requires massive datasets and sophisticated analytical approaches. Traditional studies often lack the power to detect small genetic effects or generalize findings across diverse populations. ENIGMA aims to overcome these limitations by pooling data and harmonizing methods across a global network of researchers.

Methods

ENIGMA employs a variety of advanced neuroimaging and genetic analysis techniques. Key methods include:

  • Genome-Wide Association Studies (GWAS): Analyzing millions of genetic markers (SNPs) across the genome to identify variants associated with brain measures or disease risk. This is often done through meta-analysis, combining results from multiple independent studies to increase statistical power.
  • Voxel-based Morphometry (VBM) and Meta-Analysis: Analyzing brain images at the voxel level to detect differences in brain structure (e.g., volume, thickness) between groups. ENIGMA uses a "meta-morphometry" approach to align data from different sites to a common template for large-scale analysis.
  • Diffusion Tensor Imaging (DTI): Assessing white matter integrity and connectivity.
  • Machine Learning and Predictive Modeling: Developing models to predict individual outcomes (e.g., disease risk, decline) based on brain and genetic data.
  • Polygenic Risk Scores (PRS): Combining the effects of many genetic variants to estimate an individual's overall genetic predisposition to a trait or disease.

Datasets

The paper highlights the use of data from numerous large-scale, multi-site cohorts. While specific datasets are not listed as "training" or "evaluation" sets in the traditional machine learning sense, ENIGMA's strength lies in its aggregation of data from:

  • Over 500 scientists across 200 institutions in 35 countries.
  • Tens of thousands of MRI scans, with some studies involving over 100,000 individuals.
  • Genomic data from tens of thousands of individuals.
  • Data from existing consortia such as ADNI, CHARGE, and IMAGEN.

The ENIGMA-Lifespan project specifically aims to gather normative data across the human lifespan (ages 2-92).

Key Insights and Ideas

  • Subtle but Consistent Genetic Effects: ENIGMA has demonstrated that even genetic variants explaining less than 1% of brain measure variance can be detected through large-scale meta-analyses, revealing consistent effects across diverse populations.
  • Genetics and Brain Structure are Intertwined: The consortium has identified numerous genetic loci linked to variations in brain volume and structure, suggesting a significant genetic contribution to brain morphology.
  • Disease-Specific and Overlapping Brain Patterns: ENIGMA's disease studies are revealing both unique brain structural differences associated with specific disorders (e.g., hippocampus in major depression) and common patterns across multiple conditions (e.g., in schizophrenia).
  • The Power of Big Data: The sheer scale of ENIGMA's data aggregation is crucial for detecting small effects that would be missed in smaller, single-site studies.
  • Toward Precision Medicine: The ultimate goal is to use these findings to develop predictive models for individual risk assessment, diagnosis, and treatment, paving the way for personalized medicine.

Key Results

  • ENIGMA's genomic screens have revealed at least 8 genetic loci affecting brain volumes.
  • Studies on schizophrenia and major depression have found replicable disease effects on the brain that are consistent worldwide.
  • Genome-wide association studies have identified numerous genetic loci associated with brain structure and disease risk (e.g., over 100 loci for schizophrenia).
  • The ENIGMA-Lifespan project is generating vast pools of normative brain data across the lifespan.

Limitations and Future Work

  • Reproducibility Challenges: The paper acknowledges the historical challenges with reproducibility in genetics and neuroimaging, emphasizing the importance of large sample sizes and meta-analyses for robust findings.
  • Generalizability of Models: Predictive models are most effective in cohorts similar to those where discoveries were made. ENIGMA's diversification across 33 countries aims to address this.
  • Confounding Factors: The influence of age, environment, medication, disease duration, and co-morbidities can confound findings and requires careful modeling.
  • Heterogeneity of Effects: SNP effects may not be constant across all cohorts and can vary with age, environment, or other factors.
  • Future Directions:
    • Developing more refined predictive models that account for a wider range of factors.
  • Investigating the interplay between genetic and environmental factors (epigenetics).
  • Expanding analyses to include more complex imaging measures like white matter microstructure (DTI) and brain connectivity.
  • Utilizing machine learning to build more robust predictive models for individual assessment and clinical trial design.
  • Further exploring brain asymmetry and its relation to disease.

Summary updated

BibTeX
@article{member2017enigma,
  author = {Jieping Ye, IEEE Fellow \& ACM Distinguished Member and Paul M. Thompson and Ole A. Andreassen and Alejandro Arias-Vasquez and Carrie E. Bearden and Premika S. Boedhoe and Rachel M. Brouwer and Randy L. Buckner and Jan K. Buitelaar and Kazima B. Bulayeva and Dara M. Cannon and Ronald A. Cohen and Patricia J. Conrod and Anders M. Dale and Ian J. Deary and Emily L. Dennis and Marcel A. de Reus and Sylvane Desrivières and Danai Dima and Gary Donohoe and Simon E. Fisher and Jean-Paul Fouche and et al.},
  title = {{ENIGMA and the individual: Predicting factors that affect the brain in 35 countries worldwide}},
  journal = {NeuroImage},
  year = {2017},
  doi = {10.1016/j.neuroimage.2015.11.057},
}
Figure