Análisis de asociación genómica computacional y eficiente en recursos para estudios de imagen a gran escala

Área de la Ciencia:

• Neuroscience
• Genetics
• Computational Biology

Sus antecedentes:

• Imaging genetics integrates genetic variations with brain imaging data.
• High-dimensional data presents significant computational challenges in voxel-level genome-wide association studies.

Objetivo del estudio:

• Introduce a novel framework, Representation learning-based Voxel-level Genetic Analysis (RVGA), to address computational burdens.
• Enhance statistical power and enable comprehensive genetic analyses of brain imaging data.

Principales métodos:

• Developed RVGA, a representation learning-based framework for voxel-level genetic analysis.
• RVGA reduces computational time and storage by over 200 times.
• Implemented a unified estimator for voxel heritability and genetic correlations.

Principales resultados:

• Applied RVGA to UK Biobank data (n=53,454) for hippocampus shape and white matter microstructure.
• Identified 39 novel loci for hippocampus shape and 275 for white matter microstructure.
• Discovered genetic correlations between brain regions and phenotypes like educational attainment and schizophrenia.

Conclusiones:

• RVGA offers a computationally efficient solution for large-scale imaging genetics studies.
• The framework facilitates the discovery of novel genetic associations and understanding of brain-phenotype relationships.
• RVGA replicates known associations and uncovers new genetic insights into brain structure and function.