Modelling allele-specific expression at single-cell resolution

Genetic variation plays a crucial role in shaping gene regulatory programs. These molecular programs govern how cells acquire unique identities through precise spatial and temporal control of gene activity. Mutations at cis-regulatory elements affect transcription factors binding, change the timing and dosage of gene expression and lead to phenotypical divergence. Allele-specific expression (ASE) in F1 hybrids offers a powerful approach to studying regulatory variation. Unlike direct gene expression comparisons between isogenic strains, F1 hybrid analysis avoids confounding effects from cell composition, developmental stage and environmental factors. Additionally, as both alleles reside in the same nuclear environment and are exposed to the same trans-factors, the difference in ASE reveals the effects of cis-regulatory variation. Single-cell RNA sequencing (scRNA-seq) has been instrumental in uncovering lineage-specific and transient forms of ASE, revealing regulatory complexity at a scale not feasible with bulk sequencing. Single-cell data analysis provides cell type-specific resolution but also presents technical challenges introduced by low read coverage and drop-out events. These issues are further aggravated when analysing allele-specific counts, as the signal is split between two gene copies. In my work, I used single-cell technologies to assess the impact of natural genetic variation during development, using mouse cortical organoids as a model system. We started by developing a computational pipeline for accurate allelic reads mapping and quantification, which can be applied to various sequencing technologies. Next, we built a statistical framework, ASPEN, to detect allelic imbalance via empirical Bayes-moderated beta-binomial models optimised for single-cell data. Unlike existing methods, ASPEN evaluates changes in allelic variation, including those on a temporal scale, that might reflect an interesting biological phenomenon such as transient gene activation. We applied ASPEN to the brain organoid data generated from a panel of four F1 mouse hybrids, spanning different levels of evolutionary divergence between the parental strains. This analysis represents a comprehensive map of cis-acting transcriptional regulatory variation across developing cortical cell types. Further, it revealed gene populations undergoing dynamic changes in allelic distribution during cortical neurogenesis underpinning changes in cortical structure and function.