Decoding Brain Somatic Mosaicism with New Single-Cell Copy Number Analysis Methods

Abstract

Copy number variants (CNVs) represent a significant but understudied form of somatic variation in the human brain, with potential implications for neurodevelopment, aging and disease. While single-cell whole-genome sequencing (scWGS) enables genome-wide profiling at single-cell resolution, existing computational methods struggle to accurately detect non-clonal CNVs, limiting our understanding of genomic mosaicism in the brain. In this thesis, I present two novel and complementary computational approaches for high-resolution CNV analysis in single cells. The first, HiScanner, is a CNV detection method that integrates single-cell assay-specific characteristics and introduces innovations in bin size optimization, read depth normalization, and joint segmentation across cells. Through extensive benchmarking experiments, I demonstrate HiScanner’s superior performance compared to existing tools. The second is a validation method that leverages unique molecular patterns from tagmentation-based scWGS, representing the first tool that exploits fragment overlap patterns to corroborate CNV predictions. I then apply these tools to investigate CNVs in three biological contexts: tumor evolution in paired initial and recurrent meningiomas, age-related genomic changes in neurotypical human brains, and developmental patterns in fetal and postnatal brain tissues. By analyzing both scWGS and multimodal single-cell data (paired RNA-seq and ATAC-seq), I characterize cell-type-specific CNV patterns and their potential functional implications. This work establishes a robust framework for studying somatic CNVs at single-cell resolution and provides insights into genomic instability in brain development, aging, and disease.