Abstract: In the dawning era of large-scale biomedical data, multidimensional phenotype vectors will play an increasing role in examining the genetic underpinnings of brain features, behaviour and disease. For example, shape measurements derived from brain MRI scans are multidimensional geometric descriptions of brain structure and provide an alternate class of phenotypes that remains largely unexplored in genetic studies. Here we extend the concept of heritability to multidimensional traits, and present the first comprehensive analysis of the heritability of neuroanatomical shape measurements across an ensemble of brain structures based on genome-wide SNP and MRI data from 1,320 unrelated, young and healthy individuals. We replicate our findings in an extended twin sample from the Human Connectome Project (HCP). Our results demonstrate that neuroanatomical shape can be significantly heritable, above and beyond volume, and can serve as a complementary phenotype to study the genetic determinants and clinical relevance of brain structure.
Introduction
In the field of neuroimaging genetics, researchers typically look for correlations between genetic variants and "univariate" traits—single numbers like the total volume of the hippocampus. However, volume is a "lossy" compression of anatomy. Two individuals might have the same hippocampal volume but vastly different shapes, perhaps due to localized changes in sub-nuclei or different mechanical tensions during brain development.
The challenge has been that "shape" is hard to define mathematically as a single number; it is usually represented as a high-dimensional vector. Before this study, there was no standard, computationally efficient way to estimate the SNP heritability (the proportion of variation explained by common genetic variants) for these complex, multidimensional vectors in large samples of unrelated individuals.
Methods
The researchers combined advanced geometry with multivariate statistics to create a robust analytical pipeline:
- Shape-DNA: To describe shape, they used the Laplace-Beltrami Spectrum (LBS), also known as "Shape-DNA." This method treats a brain structure's surface like a drumhead and calculates its "intrinsic frequencies." This yields a multidimensional descriptor that is "isometry-invariant"—meaning it doesn't change if the brain is rotated or shifted in the scanner.
- Multivariate Heritability Estimator: They extended the classical Haseman-Elston regression to the multivariate case. This allowed them to estimate heritability for a vector of traits without the massive computational burden of traditional likelihood-based models.
- Visualization: Because a list of eigenvalues is hard to visualize, they used Principal Component Analysis (PCA) to identify the "primary modes" of shape variation and mapped these back onto 3D brain models to show exactly how the structures deform.
Datasets
The study utilized three primary datasets:
- Discovery Set (GSP): The Brain Genomics Superstruct Project, consisting of 1,320 unrelated, young, and healthy individuals for SNP heritability calculation.
- Replication Set (HCP): The Human Connectome Project, involving 590 individuals (including twins) to replicate findings using classical twin-based models.
- Reliability Set: A subset of 42 subjects with repeated scans to assess test-retest stability.
Key Results
The analysis revealed that neuroanatomical shape is a highly sensitive and heritable phenotype. Structures including the caudate, cerebellum, corpus callosum, hippocampus, third ventricle, and putamen all exhibited shape heritability estimates ($h^2_{SNP}$) greater than 0.25.
- Shape vs. Volume: In the GSP sample, shape heritability reached FDR-corrected significance for several structures, whereas volume heritability for those same structures did not reach significance in the same sample size. This suggests that shape captures unique genetic signals that volume-based studies miss.
- Statistical Power: Surprisingly, multidimensional traits offer more statistical power. The authors proved that their multivariate estimator has reduced uncertainty compared to univariate estimates because it aggregates signal across the entire shape descriptor.
- Genetic Morphology: By visualizing the first principal component of shape, the authors could show exactly what the "genetic" version of a brain structure looks like. For example, the primary genetic variation in the caudate involves a trade-off between a larger "head" and a shorter "tail."
Discussion and Future Work
While shape descriptors offer a "higher-resolution" phenotype, they are also more sensitive to image segmentation errors than volume measurements. Improving automated brain mapping will be crucial for future studies. Furthermore, while 1,320 subjects is excellent for heritability estimation, much larger cohorts will be needed to perform a full Genome-Wide Association Study (GWAS) to identify the specific genetic variants that control these shapes. Future research may also investigate the heritability of brain asymmetry—whether the genes shaping the left side of the brain differ from those shaping the right.
Summary updated