Petrova, V.; Niu, M.; Vierbuchen, T.; Wong, E. S.

The use of F1 hybrids, the offspring of inbred parental strains, is a powerful strategy for detecting cis- and trans-regulatory contributions to transcription. At single-cell resolution, this system enables unique insights into a diversity of transcriptional phenomena. However, the detection of allelic imbalance and variance in single-cell expression data is limited by extreme sparsity and overdispersion of the count data. Here, we present ASPEN, a statistical framework for robustly modelling cis-regulatory variation in single-cell RNA-sequencing data from F1 hybrids. ASPEN integrates a sensitive allelic quantification pipeline with an adaptive shrinkage approach, facilitating accurate inference of both mean allelic imbalance and variance without over-regularizing stably expressed genes. Through extensive simulation, we show that ASPEN reliably identifies cell state-specific allelic imbalances, even under sparse data conditions. Applying ASPEN to F1 hybrid mouse brain organoids and antigen-activated T cells, we identified 2,997 genes exhibiting significant allelic imbalance. In T cells, ASPEN revealed dynamic changes in allelic variance in 59 genes, including key transcriptional regulators, kinases, and components of immune signaling pathways. ASPEN also detected random monoallelic expression events and incomplete X-inactivation in neuronal progenitors. Together, these findings underscore ASPEN\'s capacity to elucidate the regulatory architecture of gene expression at allelic resolution.