Single-cell genomics is a powerful tool for studying heterogeneous tissues such as the brain. Yet little is understood about how genetic variants influence cell-level gene expression. Addressing this, we uniformly processed single-nuclei, multiomics datasets into a resource comprising >2.8 million nuclei from the prefrontal cortex across 388 individuals. For 28 cell types, we assessed population-level variation in expression and chromatin across gene families and drug targets. We identified >550,000 cell type–specific regulatory elements and >1.4 million single-cell expression quantitative trait loci, which we used to build cell-type regulatory and cell-to-cell communication networks. These networks manifest cellular changes in aging and neuropsychiatric disorders. We further constructed an integrative model accurately imputing single-cell expression and simulating perturbations; the model prioritized ~250 disease-risk genes and drug targets with associated cell types.

Multimodal single-cell sequencing technologies provide unprecedented information on cellular heterogeneity from multiple layers of genomic readouts. However, joint analysis of two modalities without properly handling the noise often leads to overfitting of one modality by the other and worse clustering results than vanilla single-modality analysis. How to efficiently utilize the extra information from single cell multi-omics to delineate cell states and identify meaningful signal remains as a significant computational challenge. In this work, we propose a deep learning framework, named SAILERX, for efficient, robust, and flexible analysis of multi-modal single-cell data. SAILERX consists of a variational autoencoder with invariant representation learning to correct technical noises from sequencing process, and a multimodal data alignment mechanism to integrate information from different modalities. Instead of performing hard alignment by projecting both modalities to a shared latent space, SAILERX encourages the local structures of two modalities measured by pairwise similarities to be similar. This strategy is more robust against overfitting of noises, which facilitates various downstream analysis such as clustering, imputation, and marker gene detection. Furthermore, the invariant representation learning part enables SAILERX to perform integrative analysis on both multi- and single-modal datasets, making it an applicable and scalable tool for more general scenarios.

Recent advances in single-cell sequencing technologies have provided unprecedented opportunities to measure the gene expression profile and RNA velocity of individual cells. However, modeling transcriptional dynamics is computationally challenging because of the high-dimensional, sparse nature of the single-cell gene expression measurements and the nonlinear regulatory relationships. Here, we present DeepVelo , a neural network–based ordinary differential equation that can model complex transcriptome dynamics by describing continuous-time gene expression changes within individual cells. We apply DeepVelo to public datasets from different sequencing platforms to (i) formulate transcriptome dynamics on different time scales, (ii) measure the instability of cell states, and (iii) identify developmental driver genes via perturbation analysis. Benchmarking against the state-of-the-art methods shows that DeepVelo can learn a more accurate representation of the velocity field. Furthermore, our perturbation studies reveal that single-cell dynamical systems could exhibit chaotic properties. In summary, DeepVelo allows data-driven discoveries of differential equations that delineate single-cell transcriptome dynamics.

The emergence of single-cell multiomics data provides unprecedented opportunities to scrutinize the transcriptional regulatory mechanisms controlling cell identity. However, how to use those datasets to dissect the cis-regulatory element (CRE)–to–gene relationships at a single-cell level remains a major challenge. Here, we present DIRECT-NET, a machine-learning method based on gradient boosting, to identify genome-wide CREs and their relationship to target genes, either from parallel single-cell gene expression and chromatin accessibility data or from single-cell chromatin accessibility data alone. By extensively evaluating and characterizing DIRECT-NET’s predicted CREs using independent functional genomics data, we find that DIRECT-NET substantially improves the accuracy of inferring CRE-to-gene relationships in comparison to existing methods. DIRECT-NET is also capable of revealing cell subpopulation–specific and dynamic regulatory linkages. Overall, DIRECT-NET provides an efficient tool for predicting transcriptional regulation codes from single-cell multiomics data.

Post-traumatic stress disorder (PTSD) is a polygenic disorder occurring after extreme trauma exposure. Recent studies have begun to detail the molecular biology of PTSD. However, given the array of PTSD-perturbed molecular pathways identified so far¹, it is implausible that a single cell type is responsible. Here we profile the molecular responses in over two million nuclei from the dorsolateral prefrontal cortex of 111 human brains, collected post-mortem from individuals with and without PTSD and major depressive disorder. We identify neuronal and non-neuronal cell-type clusters, gene expression changes and transcriptional regulators, and map the epigenomic regulome of PTSD in a cell-type-specific manner. Our analysis revealed PTSD-associated gene alterations in inhibitory neurons, endothelial cells and microglia and uncovered genes and pathways associated with glucocorticoid signalling, GABAergic transmission and neuroinflammation. We further validated these findings using cell-type-specific spatial transcriptomics, confirming disruption of key genes such as SST and FKBP5. By integrating genetic, transcriptomic and epigenetic data, we uncovered the regulatory mechanisms of credible variants that disrupt PTSD genes, including ELFN1, MAD1L1 and KCNIP4, in a cell-type-specific context. Together, these findings provide a comprehensive characterization of the cell-specific molecular regulatory mechanisms that underlie the persisting effects of traumatic stress response on the human prefrontal cortex.

Complex biological tissues consist of numerous cells in a highly coordinated manner and carry out various biological functions. Therefore, segmenting a tissue into spatial and functional domains is critically important for understanding and controlling the biological functions. The emerging spatial transcriptomics technologies allow simultaneous measurements of thousands of genes with precise spatial information, providing an unprecedented opportunity for dissecting biological tissues. However, how to utilize such noisy, sparse, and high dimensional data for tissue segmentation remains a major challenge. Here, we develop a deep learning-based method, named SCAN-IT by transforming the spatial domain identification problem into an image segmentation problem, with cells mimicking pixels and expression values of genes within a cell representing the color channels. Specifically, SCAN-IT relies on geometric modeling, graph neural networks, and an informatics approach, DeepGraphInfomax. We demonstrate that SCAN-IT can handle datasets from a wide range of spatial transcriptomics techniques, including the ones with high spatial resolution but low gene coverage as well as those with low spatial resolution but high gene coverage. We show that SCAN-IT outperforms state-of-the-art methods using a benchmark dataset with ground truth domain annotations.