Abstract:We propose genomic principal and independent component analysis (PCA, ICA) to decompose a large set of univariate genome‐wide association study (GWAS) statistics of multimodal brain traits into more interpretable latent genomic components. Our results indicate that genomic PCA and ICA decompose genetic effects on imaging derived phenotypes from GWAS statistics with high reproducibility by taking advantage of the inherent pleiotropic patterns. Genomic components clustered along the boundaries of neuroimaging modalities. These findings encourage further applications of genomic PCA and ICA as fully data‐driven methods to effectively reduce the dimensionality, enhance the signal to noise ratio and improve interpretability of high‐dimensional multitrait genome‐wide analyses. The highly polygenic and pleiotropic nature of behavioural traits, psychiatric disorders and structural and functional brain phenotypes complicate mechanistic interpretation of related genome‐wide association study (GWAS) signals, thereby obscuring underlying causal biological processes. We propose genomic principal and independent component analysis (PCA, ICA) to decompose a large set of univariate GWAS statistics of multimodal brain traits into more interpretable latent genomic components. Here we introduce and evaluate this novel methods various analytic parameters and reproducibility across independent samples. Two UK Biobank GWAS summary statistic releases of 2240 imaging‐derived phenotypes (IDPs) were retrieved. Genome‐wide beta‐values and their corresponding standard‐error scaled z‐values were decomposed using genomic PCA/ICA. We evaluated variance explained at multiple dimensions up to 200. We tested the inter‐sample reproducibility of output of dimensions 5, 10, 25 and 50. Reproducibility statistics of the respective univariate GWAS served as benchmarks. Reproducibility of 10‐dimensional PCs and ICs showed the best trade‐off between model complexity and robustness and variance explained (PCs: |rz − max| = 0.33, |rraw − max| = 0.30; ICs: |rz − max| = 0.23, |rraw − max| = 0.19). Genomic PC and IC reproducibility improved substantially relative to mean univariate GWAS reproducibility up to dimension 10. Genomic components clustered along neuroimaging modalities. Our results indicate that genomic PCA and ICA decompose genetic effects on IDPs from GWAS statistics with high reproducibility by taking advantage of the inherent pleiotropic patterns. These findings encourage further applications of genomic PCA and ICA as fully data‐driven methods to effectively reduce the dimensionality, enhance the signal to noise ratio and improve interpretability of high‐dimensional multitrait genome‐wide analyses.

AC-PCA Adjusts for Confounding Variation and Recovers the Anatomical Structure of Neocortex

Simultaneous Dimension Reduction And Adjustment For Confounding Variation

AC-PCoA: Adjustment for Confounding Factors Using Principal Coordinate Analysis

Joint Modeling of Cellular Heterogeneity and Condition Effects with scPCA in Single-Cell RNA-Seq

Demixed principal component analysis of neural population data

Principal and independent genomic components of brain structure and function

Editorial: Advances in Rare Earth Resource Utilization and Functional Materials

Exploring Brain Dynamic Functional Connectivity Using Improved Principal Components Analysis Based on Template Matching

PARE: A framework for removal of confounding effects from any distance-based dimension reduction method

Identifying associations in dense connectomes using structured kernel principal component regression

A powerful and efficient multivariate approach for voxel-level connectome-wide association studies

ACE of Space: Estimating Genetic Components of High-Dimensional Imaging Data

A Powerful and Ef Fi Cient Multivariate Approach for Voxel-Level Connectome-Wide Association Studies

Genomic loci influence patterns of structural covariance in the human brain

nPCA: a linear dimensionality reduction method using a multilayer perceptron

Identifying patterns differing between high-dimensional datasets with generalized contrastive PCA

Interpretable principal component analysis for multilevel multivariate functional data

Identifying age-specific gene signatures of the human cerebral cortex with joint analysis of transcriptomes and functional connectomes

Spatially aware dimension reduction for spatial transcriptomics

Whole human-brain mapping of single cortical neurons for profiling morphological diversity and stereotypy

Beyond out-of-sample: robust and generalizable multivariate neuroanatomical patterns of psychiatric problems in youth