Deconvolution of spatial transcriptomics data via graph contrastive learning and partial least square regression.

Auditory hair cells transduce sound to the brain, and in mammals, these cells reside together with supporting cells in the sensory epithelium of the cochlea, called the organ of Corti. To establish the organ's delicate function during development and differentiation, spatiotemporal gene expression is strictly controlled by chromatin accessibility and cell type–specific transcription factors, jointly representing the regulatory landscape. Bulk sequencing technology and cellular heterogeneity obscured investigations on the interplay between transcription factors and chromatin accessibility in inner ear development. To study the formation of the regulatory landscape in hair cells, we collected single-cell chromatin accessibility profiles accompanied by single-cell RNA data from genetically labeled murine hair cells and supporting cells after birth. Using an integrative approach, we predicted cell type–specific activating and repressing functions of developmental transcription factors. Furthermore, by integrating gene expression and chromatin accessibility data sets, we reconstructed gene regulatory networks. Then, using a comparative approach, 20 hair cell–specific activators and repressors, including putative downstream target genes, were identified. Clustering of target genes resolved groups of related transcription factors and was used to infer their developmental functions. Finally, the heterogeneity in the single-cell data allowed us to spatially reconstruct transcriptional as well as chromatin accessibility trajectories, indicating that gradual changes in the chromatin accessibility landscape are lagging behind the transcriptional identity of hair cells along the organ's longitudinal axis. Overall, this study provides a strategy to spatially reconstruct the formation of a lineage-specific regulatory landscape using a single-cell multi-omics approach.

Recent extensions of single-cell studies to multiple data modalities raise new questions regarding experimental design. For example, the challenge of sparsity in single-omics data might be partly resolved by compensating for missing information across modalities. In particular, deep learning approaches, such as deep generative models (DGMs), can potentially uncover complex patterns via a joint embedding. Yet, this also raises the question of sample size requirements for identifying such patterns from single-cell multi-omics data. Here, we empirically examine the quality of DGM-based integrations for varying sample sizes. We first review the existing literature and give a short overview of deep learning methods for multi-omics integration. Next, we consider eight popular tools in more detail and examine their robustness to different cell numbers, covering two of the most common multi-omics types currently favored. Specifically, we use data featuring simultaneous gene expression measurements at the RNA level and protein abundance measurements for cell surface proteins (CITE-seq), as well as data where chromatin accessibility and RNA expression are measured in thousands of cells (10x Multiome). We examine the ability of the methods to learn joint embeddings based on biological and technical metrics. Finally, we provide recommendations for the design of multi-omics experiments and discuss potential future developments.

Single-cell multi-omics sequencing techniques have rapidly developed in the past few years. Clustering analysis with single-cell multi-omics data may give us novel perspectives to dissect cellular heterogeneity. However, multi-omics data have the properties of inherited large dimension, high sparsity and existence of doublets. Moreover, representations of different omics from even the same cell follow diverse distributions. Without proper distribution alignment techniques, clustering methods will encounter less separable clusters easily affected by less informative omics data. We developed MoClust, a novel joint clustering framework that can be applied to several types of single-cell multi-omics data. A selective automatic doublet detection module that can identify and filter out doublets is introduced in the pretraining stage to improve data quality. Omics-specific autoencoders are introduced to characterize the multi-omics data. A contrastive learning way of distribution alignment is adopted to adaptively fuse omics representations into an omics-invariant representation. This novel way of alignment boosts the compactness and separableness of clusters, while accurately weighting the contribution of each omics to the clustering object. Extensive experiments, over both simulated and real multi-omics datasets, demonstrated the powerful alignment, doublet detection and clustering ability features of MoClust. An implementation of MoClust is available from https://doi.org/10.5281/zenodo.7306504. Supplementary data are available at Bioinformatics online.

stSCI: A multi-task learning framework for integrative analysis of single-cell and spatial transcriptomics data

Recent advancements in single-cell multi-omics technologies have significantly improved our ability to explore cellular heterogeneity at an unprecedented resolution. These innovations enable the simultaneous profiling of genomic, transcriptomic, proteomic, and epigenetic data at the single-cell level, providing comprehensive insights into cellular states and their regulatory mechanisms. However, integrating multi-omics data remains challenging due to its high dimensionality, technical noise, and biological complexity. To address these challenges, we introduce scWDAC (single-cell weighted distance adaptive clustering), a novel clustering method for single-cell multi-omics data. scWDAC utilizes a weighted distance penalty and adaptive graph regularization to effectively integrate multiple omics layers. Key innovations of scWDAC include using a weighted distance penalty to capture cell-to-cell similarities across different omics modalities, and applying adaptive graph regularization on a consensus matrix to enforce cross-modal consistency. The framework optimizes both global consistency and local accuracy, ensuring a robust exploration of cellular structures across all omics layers. The effectiveness of scWDAC is evaluated through extensive experiments on ten paired single-cell multi-omics datasets. The results demonstrate that scWDAC outperforms existing clustering methods in terms of clustering accuracy, robustness to noise, and biological interpretability.

Enhancers play an important role in transcriptional regulation by modulating gene expression from distal genomic locations. Although single-cell ATAC and RNA sequencing (scATAC/RNA-seq) data have been leveraged to infer enhancer-gene links, establishing regulatory links between enhancers and their target genes remains a challenge due to the absence of chromatin conformation information. Here, we present SCEG-HiC, a machine learning method based on weighted graphical lasso, which decodes enhancer-gene links from single-cell multi-omics data by integrating bulk average Hi-C as prior knowledge. SCEG-HiC supports both paired scATAC/RNA-seq and scATAC-only inputs, improving prediction accuracy while retaining context-specific correlations and enabling the discovery of biologically relevant links. Comprehensive evaluation across 10 human and mouse single-cell multi-omics datasets shows that SCEG-HiC outperforms existing single-cell models. Application of SCEG-HiC to COVID-19 datasets illustrates its capacity to more reliably reconstruct gene regulatory networks underlying disease severity, and elucidate functional associations between noncoding variants and their putative target genes. SCEG-HiC is freely available as an open-source and user-friendly R package, facilitating broad applications in regulatory genomics research.

Novel cellular therapy has the potential to change cancer treatment. However, to design new therapies requires a better understanding of the tumor microenvironment (TME) and the cellular dynamics that occur within the tumor. This requires a high resolution and subcellular view of cancer cells, immune cells while identifying the neighborhood architecture that exists between them. Spatial transcriptomics and spatial proteomics show subcellular information and can be used to identify cellular cell types and find spatial deviation between cells. However, there are a lack of tools to investigate cellular behaviors on an individual cell basis while integrating pathological information, limiting the future of spatial biology. We introduce CellClique, a tool that assists in dissecting TMEs using spatial transcriptomics with the Xenium platform and Hematoxylin and Eosin (H&E) stains. CellClique integrates both generative and analytical capabilities to provide a comprehensive view of TMEs. On the generative end, CellClique uses generative AI to model both gene expressions and cell-neighborhood interactions for cell annotation, gene expression prediction, and gene imputation. On the analytical end, CellClique offers a comprehensive four-way spatial analysis scheme aimed at 1) understanding the effects of neighboring cells on a target cell, 2) dissecting cancer-normal cell boundaries, 3) observing immune invasion mechanisms, and 4) exploring cellular subtypes with the TME. We applied CellClique to study the cellular landscape in lung and colon cancers using multiple different neighborhood analyses. We generated proximal (10 microns), local (60 microns), and distal neighborhoods (120 microns) for each cell within each tissue and found distinct expression profiles of tumor-infiltrated lymphocytes, which are functionally different from lymphocytes excluded from the tumor. Additionally, we discovered expression variation patterns in cancer-normal boundary cell communities with various cell compositions. These compositions changed significantly as distance from the boundary changed. We also compared lymphocytes at different stages of life by analyzing expression differences among lymphocytes classified with over 80% confidence, those with a 50% probability of being lymphocytes, and those with below 20% probability. We further compared how these neighborhoods changed between lung cancer and colon cancer and showed cancer specific spatial deviations and expression. Together, these methods paint a complex picture of the TME but also identifies numerous avenues for future discovery and molecular investigation. These new approaches will guide future cellular therapy and the development of new treatments. The future integration of AI and machine learning will continue to make a tremendous impact on spatial transcriptomics and our approach to treating disease. Citation Format: Sachini Weerasekara, Natasha Darras, Nicolas Fernandez, Melinda Chen, Alina Ainbinder, Colles Price. CellClique: Dissecting tumor microenvironments at the single cell level using generative AI and spatial transcriptomics [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2418.

Interpretable data integration for single-cell and spatial multi-omics.

Recent advances in single-cell multi-omics technologies have provided unprecedented insights into regulatory processes. We introduce TREASMO, a versatile Python package designed to quantify and visualize transcriptional regulatory dynamics in single-cell multi-omics datasets. TREASMO has four modules, spanning data preparation, correlation quantification, downstream analysis and visualization, enabling comprehensive dataset exploration. By introducing a novel single-cell gene-peak correlation strength index, TREASMO facilitates accurate identification of regulatory changes at single-cell resolution. Validation on a hematopoietic stem and progenitor cell dataset showcases TREASMO's capacity in quantifying the gene-peak correlation strength at the single-cell level, identifying regulatory markers and discovering temporal regulatory patterns along the trajectory.

My talk will present three computational frameworks we developed to study cytokine signaling activities and cell-cell communications during the antitumor immune response, using tumor single-cell and spatial transcriptomics. The basic immunology tool to study cytokine signaling mostly measures cytokine release, which is transient and does not represent downstream target activities. Therefore, we first developed the CytoSig platform, providing a database of target genes modulated by cytokines and a predictive model of cytokine signaling cascades from transcriptomic profiles. We collected 20,591 transcriptome profiles for human cytokine, chemokine, and growth factor responses. This atlas of transcriptional patterns induced by cytokines enabled the reliable prediction of signaling activities in distinct cell populations in infectious diseases, chronic inflammation, and cancer using bulk and single-cell transcriptomic data. CytoSig revealed previously unidentified roles of many cytokines, such as BMP6 as an anti-inflammatory factor. Then, based on CytoSig, we developed Tres, a computational model utilizing single-cell transcriptomic data to identify signatures of T cells that are resilient to immunosuppressive signals, such as TGF-β1, TRAIL, and prostaglandin E2. Tres reliably predicts clinical responses to immunotherapy in multiple cancer types using bulk T cell transcriptomic data from pre-treatment patient tumors or infusion/pre-manufacture samples for cellular immunotherapies. Further, Tres identified FIBP as a candidate immunotherapy target to potentiate adoptive cell therapy efficacies. FIBP knockout in T cells enhanced adoptive cell therapy by down-regulating T cells' cholesterol metabolism. Last, I will briefly show our SpaCET framework for deconvolving cell fractions and identifying cell-cell interactions in tumor spatial transcriptomics data. SpaCET resolved several challenges in spatial transcriptomics analysis that previous methods did not address sufficiently. Through coupling cell fractions with ligand-receptor co-expression analysis, SpaCET reveals that intercellular interactions tend to be located at the tumor-immune boundaries. Citation Format: Peng Jiang. Inference of intercellular signaling activities in tumor spatial and single-cell transcriptomics, with applications in identifying cancer immunotherapy targets [abstract]. In: Proceedings of the AACR-NCI-EORTC Virtual International Conference on Molecular Targets and Cancer Therapeutics; 2023 Oct 11-15; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2023;22(12 Suppl):Abstract nr IA002.

Cancer-associated fibroblasts (CAFs) are a diverse group of cells that significantly contribute to reshaping the tumor microenvironment (TME), and no research has systematically explored the molecular landscapes of senescence related CAFs (senes CAF) in NB. We utilized pan-cancer single cell and spatial transcriptomics analysis to identify the subpopulation of senes CAFs via senescence related genes, exploring its spatial distribution characteristics. Harnessing the maker genes with prognostic significance, we delineated the molecular landscapes of senes CAFs in bulk-seq data. We established the senes CAFs related signature (SCRS) by amalgamating 12 and 10 distinct machine learning (ML) algorithms to precisely diagnose stage 4 NB and to predict prognosis in NB. Based on risk scores calculated by prognostic SCRS, patients were categorized into high and low risk groups according to median risk score. We conducted comprehensive analysis between two risk groups, in terms of clinical applications, immune microenvironment, somatic mutations, immunotherapy, chemotherapy and single cell level. Ultimately, we explore the biological function of the hub gene JAK1 in pan-cancer multi-omics landscape. Through integrated analysis of pan-cancer spatial and single-cell transcriptomics data, we identified distinct functional subgroups of CAFs and characterized their spatial distribution patterns. With marker genes of senes CAF and leave-one-out cross-validation, we selected RF algorithm to establish diagnostic SCRS, and SuperPC algorithm to develop prognostic SCRS. SCRS demonstrated a stable predictive capability, outperforming the previously published NB signatures and clinic variables. We stratified NB patients into high and low risk group, which showed the low-risk group with a superior survival outcome, an abundant immune infiltration, a different mutation landscape, and an enhanced sensitivity to immunotherapy. Single cell analysis reveals biologically cellular variations underlying model genes of SCRS. Spatial transcriptomics delineated the molecular variant expressions of hub gene JAK1 in malignant cells across cancers, while immunohistochemistry validated the differential protein levels of JAK1 in NB. Based on multi-omics analysis and ML algorithms, we successfully developed the SCRS to enable accurate diagnosis and prognostic stratification in NB, which shed light on molecular landscapes of senes CAF and clinical utilization of SCRS.

Pancreatic Cancer Ductal Adenocarcinoma (PDAC) is one of the deadliest cancers with 5-year survival of 11%. Understanding the intratumor heterogeneity is a pivotal piece to unravel the complexity of PDAC. While single cell RNASeq identifies the heterogeneous cell populations within the tumor tissue, spatially characterizing the transcriptomic profile of neoplastic and pre-neoplastic populations within the tissue remains a challenge, as the spatial dimension is usually lost upon tissue dissociation. We identified an approach to integrate spatial transcriptomics data with single cell RNASeq data. To characterize the cell populations within the tissue we performed single cell RNASeq on disease pathology-free pancreas tissue and primary PDAC samples. We profiled the transcriptomic profile of Acinar, Ductal, Acinar-to-Ductal (ADM), and Pancreatic Intraepithelial Neoplasia (PanINs) regions of interest (ROIs) across the tissue using the Nanostring GeoMx platform. Differential gene expression analysis using linear mixed-effect models of the cell-type specific ROIs defined pan-marker gene sets for each cell type, which were mapped to UMAP projections of single cell RNA sequencing data using AUCell scoring. As expected, the acinar pan-markers gene set derived from the spatial transcriptomics mapped to the manually annotated acinar population in the single cell data. On the other hand, Ductal, ADM, and PanIN pan-marker gene sets were mapped to distinct clusters that previously were not well-defined by single cell sequencing. The analysis coupled with orthogonal validation using RNAScope revealed gene signatures uniquely specific to ADM lesions and PanINs, respectively. Interestingly, our list included known markers as well as novel findings, supporting the validity of the findings. Furthermore, RNA velocity analysis using scVelo revealed a trajectory of cell evolution originating from acinar cells passing through the newly-defined ADM population and ending towards the ductal population derived from tumor samples. Overall, this integration approach of spatial and single cell transcriptomics can further define the characteristics that differentiate neoplastic and pre-neoplastic populations, as well as the potential drivers for tumorigenesis that could be therapeutically targeted. Citation Format: Ahmed M. Elhossiny, Eileen Carpenter, Padma Kadiyala, Yaqing Zhang, Filip Bednar, Arvind Rao, Timothy Frankel, Marina Pasca Di Magliano. Integrating single cell and spatial transcriptomics define gene signature for pancreatic ductal adenocarcinoma pre-neoplastic lesion [abstract]. In: Proceedings of the AACR Special Conference on Pancreatic Cancer; 2022 Sep 13-16; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2022;82(22 Suppl):Abstract nr A006.

Advances in single-cell technology enable large-scale generation of omics data, promising for clarifying gene regulatory networks governing different cell type/states. Nonetheless, prevailing methods fail to account for universal and reusable regulatory modules in GRNs, which are fundamental underpinnings of cell type landscape. We introduce cRegulon to infer regulatory modules by modeling combinatorial regulation of transcription factors based on diverse GRNs from single-cell multi-omics data. Through benchmarking and applications using simulated datasets and real datasets, cRegulon outperforms existing approaches in identifying TF combinatorial modules as regulatory units and annotating cell types. cRegulon offers new insights and methodology into combinatorial regulation.

Background: Effective cancer immunotherapy is significantly affected by intratumoral heterogeneity, not only in different cell types and functions but also in the spatial pattern of immune cells within the tumor tissue. While most single-cell RNA sequencing technologies could enable the exploration of gene expression heterogeneity at the single-cell level, they cannot reveal spatial information within the tumor for us to map the whole transcriptome with morphological context. Spatial transcriptomics is an in situ capturing technique, which profiles gene expression at the RNA level, whilst preserving the spatial information of histological tissue sections. Thus we try to develop our spatial transcriptomics assay through a serial of validation experiments and to explore the mechanism of immune therapy against tumors. Material and Method: Firstly, the tumor was frozen and sections were cryosectioned at 10 μm thickness. Sections processing including H&E staining, tissue optimization, permeabilization, reverse transcription, and cDNA library preparation was carried out following the user guide of Visium Spatial Gene Expression Reagent Kits (10X Genomics). And the final cDNA libraries were sequenced using the Illumina Novaseq instrument. Using gene expression matrix and aligned H&E image processed by Cell Ranger and Loupe Browser with default parameters, we performed data QC, normalization, dimensionality reduction, and clustering by Seurat and cell components identification in tumor microenvironment based on a multistep approach. Differential gene expression and spatial heterogeneity were analyzed in each cluster. Results: Here we established a reliable system for spatial transcriptomics assay to elucidate the mechanism of anti-PD1 treatment in MC38 syngeneic colorectal tumor. Using this system, we constructed the gene expression spatial patterns of tumor cells and immune cells, including periphery and center of tumor cell, immune-cell enrichment area, and tumor cell surrounding the immune cell. Furthermore, we explored the cell-cell interactions in which the tumor cells directly interact with adjacent non-tumor cells, for example, macrophages. Using this system, we tried to investigate the mechanism of anti-PD-1 therapy against the tumors by comparing the spatial transcriptome profiles after the treatment with anti-PD-1 or isotype control in colorectal cancer models. Interestingly, changes in both gene-level and spatial levels were observed, and it will help reveal how the immune therapy may shake up the complex ecosystem. Summary: We established a reliable system for spatial transcriptomics analysis and explored the mechanism of anti-PD-1 treatment against the colorectal cancer by using this system. Citation Format: Panpan Wang, Huacheng He, Qiyao Zhang, Jingjing Wang, Dan Lu, Fei Duan, Qiangqiang Fan, Qingyang Gu, Qunsheng Ji. Establishment of spatial transcriptomics assay to study the mechanism of immune therapy against tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6128.

Single-cell technologies: From research to application