Elham Azizi

In vivo single-cell CRISPR screens can resolve perturbation effects in native tissues, but few combine cell-type-restricted delivery, somatic tumor initiation, and combinatorial perturbation. Here we developed an RCAS-TVA-based single-cell CRISPR screening approach coupling oncogenic transformation with single- or dual-guide perturbation in defined neural progenitors. In glioma, tumor suppressor loss produced distinct genotype-dependent transcriptional states, including an NF1-loss-associated increase in mesenchymal programs. Introducing a fixed Cdkn2a(ARF) perturbation altered malignant-cell state composition, immune-cell abundance, and the effects of additional tumor-suppressor perturbations, including increased representation of Pten- and Rb1-deficient cells, suggestive of the relief of an ARF-dependent genetic barrier. These perturbations converged on transcriptional states resembling pharmacologic p53 activation in human glioblastoma. Across mouse tumors and human glioblastoma models, this response was associated with increased astrocyte-like state representation. These findings establish a framework for single-cell analysis of context-dependent somatic perturbations during tumor initiation in vivo.

Understanding the effects of genetic, molecular, and experimental perturbations is essential for decoding cellular mechanisms and guiding biomedical interventions. Existing computational approaches are typically designed for individual tasks, such as predicting perturbation responses, characterizing downstream transcriptional effects, or modeling perturbation combinations, and therefore do not provide a unified framework for learning interpretable perturbation representations while enabling causal analysis of downstream effects and characterization of condition-dependent responses. We present SHERLOCK, an interpretable deep generative framework for single-cell perturbation analysis that represents perturbation effects as structured interventions on a latent baseline cellular state. SHERLOCK learns correlated and sparse perturbation representations that organize genetic and pharmacological perturbations according to shared transcriptional responses. By formulating perturbations as interventions within a structural causal model, it enables counterfactual estimation of their downstream transcriptional effects under explicit identifiability assumptions. The same framework quantifies how perturbation responses vary across conditions and compositionally models combinatorial perturbations, enabling prediction of held-out combinations and classification of genetic interactions. Across genome-scale CRISPR, chemical, and spatial perturbation datasets, SHERLOCK recovers perturbation relationships concordant with known biological pathways and pharmacological properties, identifies condition-dependent responses, and predicts combinatorial perturbation effects. Together, SHERLOCK provides a unified framework for interpretable and causal analysis of perturbation effects across diverse single-cell perturbation experiments.

Allogeneic hematopoietic cell transplantation cures hematologic diseases but is limited by acute graft‑versus‑host disease. How human T cell clones drive epithelial injury remains poorly mapped. We studied 31 transplant recipients, integrating longitudinal T cell antigen receptor (TCR) profiling with single-cell RNA sequencing/TCR sequencing and spatial transcriptomics to track T cell clonal dynamics. We developed DecompTCR to resolve temporal dynamics and adapted computational tools to map clone phenotypes and niches in tissue. Our analyses revealed that cyclophosphamide selectively depletes alloreactive clones, although insufficient early expansion leads to incomplete depletion and severe disease. Severe graft‑versus‑host disease is marked by persistent expansion of alloreactive clones, rewiring of homeostatic cell types and diversification of donor-derived CD8⁺ clonotypes that acquire Hobit (ZNF683)⁺ tissue‑resident memory T (T_RM) cell programs during migration to epithelium. Spatial deconvolution identified CD8⁺ effector/Hobit⁺ T_RM hubs near intestinal stem‑cell-rich crypt bases and crypt‑loss regions. This clonotype‑resolved framework links tissue‑instructed T_RM cell remodeling to localized epithelial injury, nominating early-repertoire dynamics and spatial hub burden as biomarkers.

Tissues are spatially organized systems in which cell states, functions and interactions vary across spatial coordinates, forming compartments or gradients shaped by local microenvironments. Understanding how molecular features and cell–cell interactions change across space and time is central to studying development, homeostasis and disease. Addressing these questions increasingly requires the integration of multi-modal spatial data, which provides complementary views of cellular and structural organization. However, existing computational approaches typically combine modalities by weighting them equally, overlooking domain-specific technical artifacts, differences in spatial resolution and non-overlapping feature spaces. In addition, methods for spatial cell–cell communication analysis are largely developed for single-modality settings and do not model how interactions vary across the tissue. To address these gaps, we introduce LYNX, a deep generative framework that learns a shared latent representation of spatial dynamics from joint-measured modalities in the 2D or 3D domain, to provide a unified coordinate system for modeling how cell–cell interactions, phenotypes, and molecular programs vary along continuous spatial gradients. LYNX identifies spatial programs difficult to resolve with existing approaches, including metabolically coupled porto-central interaction remodeling in liver, recovery of degraded proteomic signals along the cortico-medullary axis in thymus, and branching trajectories towards DCIS and invasive niches marked by distinct stromal activation-states and immune-tumor crosstalk in breast tumor microenvironment. We demonstrate that LYNX robustly infers spatially resolved gradients, maps functional compartments and cell-cell interactions along spatial axes and is compatible across diverse spatial profiling technologies, modalities, and resolution disparities. LYNX provides a foundational and scalable framework to advance our understanding of healthy tissue physiology and to decode temporal evolution of complex diseases.

Spatially organized immune hubs of T cells and antigen-presenting cells (APCs) have been linked to immune checkpoint therapy (ICT) efficacy, yet the mechanisms underlying their function remain unclear. Using CODEX multiplex imaging, we longitudinally characterized the dynamic evolution of intratumoral cellular neighborhoods (CN) defined by triad interactions of CD4 and CD8 T cells with two distinct myeloid APC populations: cDC1s and IFN-gamma-activated macrophages. We termed this CN the immunity-promoting CN (IP-CN) and tracked its progressive development during tumor rejection induced by anti-CTLA-4/anti-PD-1 therapy. A coordinated IFN-gamma; and TNF-alpha signaling signature accompanied the IP-CN assembly. Over time, the IP-CN underwent functional maturation, forming specialized sub-neighborhoods that compartmentalized proliferating T cells at the tumor periphery versus cytotoxic T effector cells interacting with tumor cell targets. Our findings reveal a spatiotemporal mechanism by which the IP-CN sustains and amplifies cytotoxic T cell responses, demonstrating how T cell-APC neighborhoods orchestrate tumor immunity.

Multiple single-cell and spatial genomics tools have transformed our ability to deconvolve intricate diseases, including cancer. Analysis of complex, multimodal data has provided insights into genomics, cellular states and interactions in tumor ecosystems, enabling the dissection of salient biology and expanding our understanding of drug response, resistance and target discovery. However, several challenges remain before these methods can achieve their full clinical potential. Here, we discuss opportunities, barriers and potential solutions, including sample acquisition and preservation approaches, profiling methods and analytical tools for heterogeneous populations, and we provide recommendations for robust, reproducible use of these technologies in clinical settings.

Colorectal cancer (CRC) cell atlases have refined descriptive maps of tumour ecosystems, yet cross-sample integration often obscures disease-relevant patient-specific variation and remains largely correlative, limiting insight into the mechanisms and state transitions that drive progression and treatment response. Here, we develop a continual learning framework to construct a comparative single-cell CRC atlas spanning over 300 patients and 1.5 million cells, preserving inter-patient variation while aligning healthy and malignant contexts. We resolve distinct non-canonical malignant cell states, including an endoderm-like state enriched in microsatellite-stable, KRAS -mutant CRC with features of oncofetal plasticity. Cell states are recapitulated in patient-derived organoids, establishing a tractable model of reprogramming. By linking the observational atlas to a large-scale perturbation atlas using relative representations, we map perturbations that drive cells toward defined phenotypic extremes. We connect cell states to therapeutic responses, showing that MAPK inhibition induces a shift away from a proliferative phenotype and converges towards a plastic, endoderm-like state. Together, this framework moves beyond static atlases to enable mechanistic modeling of cell-state regulation and causal inference toward cell-state–directed therapies.

The integration of single-cell genomics into the chemical genetics paradigm is reshaping how researchers profile drug activity, prioritize lead candidates, and uncover new therapeutic opportunities. Traditional chemical genetic approaches, though instrumental in linking compounds to cellular phenotypes, often rely on bulk measurements that obscure important cellular heterogeneity and limit insight into mechanisms of action. By contrast, single-cell technologies offer a transformative view of how compounds influence diverse cell types and states, capturing nuanced molecular responses that further our understanding of efficacy, resistance, and polypharmacology. From cancer to neurodegenerative disorders and other disease contexts, single-cell chemical profiling enables a more precise annotation of drug-induced effects, revealing differential responses across cellular subpopulations. These methods help identify both beneficial and adverse outcomes that may not be readily predicted by a compound’s structure or known targets, enhancing preclinical prioritization and supporting rational drug repurposing strategies. As these technologies mature, advances in multiplexing, multimodal profiling, and computational analysis are expanding their scalability and applicability to increasingly complex models. The resulting data-rich assays are poised to bridge critical gaps between compound screening and clinical relevance. This review highlights the evolution of chemical genomics toward single-cell resolution and outlines emerging opportunities to leverage these methods throughout the drug discovery pipeline, from early preclinical prioritization to late-stage repurposing, ultimately accelerating the development of safer, more effective therapies.

Single-cell sequencing has revolutionized our understanding of cellular heterogeneity and responses to environmental stimuli. However, mapping transcriptomic changes across diverse cell types in response to various stimuli and elucidating underlying disease mechanisms remains challenging. Here we present Squidiff, a diffusion model-based generative framework that predicts transcriptomic changes across diverse cell types in response to environmental changes. We demonstrate the robustness of Squidiff across cell differentiation, gene perturbation and drug response prediction. Through continuous denoising and semantic feature integration, Squidiff learns transient cell states and predicts high-resolution transcriptomic landscapes over time and conditions. Furthermore, we applied Squidiff to model blood vessel organoid development and cellular responses to neutron irradiation and growth factors. Our results demonstrate that Squidiff enables in silico screening of molecular landscapes and cellular state transitions, facilitating rapid hypothesis generation and providing valuable insights into the regulatory principles of cell fate decisions.

Most patients with localized pancreatic ductal adenocarcinoma (PDAC) experience recurrence after resection. Analysis of 744 patients with resected PDAC revealed that patients with initial isolated liver-metastatic recurrence (n = 100) had significantly worse overall survival than those with initial isolated lung-metastatic recurrence (n = 31). Using single-nucleus RNA sequencing in a representative cohort, we found that transcriptional profiles of primary cancer cells with liver-metastatic recurrence and lung-metastatic recurrence were correlated with those of normal liver and lung parenchymal cells, respectively, suggesting adoption of organ-specific metastatic programs at the primary site. These signatures were confirmed in transcriptomes of PDAC lung and liver metastases, primary lung and liver tumors, and organotropic PDAC xenograft models. These signatures were independent of large genomic events, and analysis of large-scale tumor profiling data showed no genetic alterations predictive of recurrence patterns. Additional analyses suggested that metastatic recurrence may be determined early in tumorigenesis and influenced by tumor-infiltrating immune cells. Thus, pre-existing cellular states within primary tumors appear to guide organ-specific metastatic relapse.