Single-cell Gene Expression Data Research Articles

Abstract 3141: Epiregulon infers single-cell transcription factor activity to dissect mechanisms of lineage plasticity and drug response

Abstract Single-cell multiomic technologies enable holistic reconstruction of cell states and hold the potential to identify drivers of tumor evolution and drug resistance. However, the data sparsity of single-cell assays can preclude accurate estimation of transcription factor activity. In addition, gene expression alone cannot capture transcription factor activity, especially in the context of drug treatment which alters transcription factor function without suppressing gene expression. To circumvent these challenges, we have developed epiregulon, a R package that constructs gene regulatory networks and infers transcription factor (TF) activity in single cells by integrating single cell gene expression, chromatin accessibility and bulk TF occupancy data. Epiregulon applies tests of independence to identify likely transcription factor - regulatory element - target gene relationships occurring at joint probabilities exceeding the expected probabilities of independent events. With epiregulon, we are able to detect lineage factor activity at enhanced sensitivity and predict perturbations more accurately than gene expression could. We further applied this tool to understand transcription factor activity to enzalutamide treatment and identify potential drivers of enzalutamide resistance. Finally, we generated a ground truth dataset using reprogram-seq, a technology that captures multi-omic profiles of single cells upon expression of defined transcription factors. Epiregulon correctly predicts NKX2-1 targets and demonstrates that NKX2-1 expression reprograms the epigenome towards a neuroendocrine state in LNCaP cells. Epiregulon is a useful tool that constructs gene regulatory networks to model the underlying gene regulation hierarchies that drive gene expression and cell states. Citation Format: Tomasz Wlodarczyk, Jenille Tan, Kerstin Seidel, Diana Wu, Shang-Yang Chen, Aleksander Chlebowski, Timothy Keyes, Yu Guo, Aaron Lun, Christopher Siebel, Shiqi Xie, Xiaosai Yao. Epiregulon infers single-cell transcription factor activity to dissect mechanisms of lineage plasticity and drug response [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3141.

Abstract 3551: Single-cell characterization of pediatric T-cell acute lymphoblastic leukemia transcriptomes

Abstract Pediatric T-cell acute lymphoblastic leukemia (T-ALL) is a rare cancer with a diverse set of causative genomic abnormalities. The variety and limited characterization of genomic mechanisms underlying this disease presents a unique challenge for the development of improved treatments for these patients. Using the 10x Genomics Chromium system, we have generated single-cell gene expression (scRNA) and chromatin accessibility (scATAC) data for blood and/or bone marrow samples collected at diagnosis from eight pediatric T-ALL patients. We sequenced 21,669 high-confidence cells and performed normalization and downstream analysis using the Seurat v4.0.2 package in R 4.0.3. Our data show that patient cancer cells show highly individual transcriptomes and are often characterized by combinatorial expression patterns of genes including MEF2C, BCL11B, HOXA9, NKX2-1, and LMO2, which have been previously implicated for their roles in various T-ALL subtypes. Additionally, we observe that each patient has a unique ratio of cancer cells with a CD4+ T-cell, CD8+ T-cell, and hematopoietic stem cell-like identity, which may be linked to treatment response and patient outcome. This dataset represents one of the first forays into exploring the biology of pediatric T-ALL at the single-cell level. Our comparative analysis of patient cancer cells reveals novel insights into the transcriptional and chromatin-related for distinct subsets of pediatric T-ALL cases, paving the way for future treatment selection and drug development to improve outcomes for these patients. Citation Format: Irina Pushel, Byunggil Yoo, Daniel Louiselle, Margaret Gibson, Midhat S. Farooqi, Keith August. Single-cell characterization of pediatric T-cell acute lymphoblastic leukemia transcriptomes. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3551.

Integrating Multiple Single-Cell RNA Sequencing Datasets Using Adversarial Autoencoders.

Single-cell RNA sequencing (RNA-seq) has been demonstrated to be a proven method for quantifying gene-expression heterogeneity and providing insight into the transcriptome at the single-cell level. When combining multiple single-cell transcriptome datasets for analysis, it is common to first correct the batch effect. Most of the state-of-the-art processing methods are unsupervised, i.e., they do not utilize single-cell cluster labeling information, which could improve the performance of batch correction methods, especially in the case of multiple cell types. To better utilize known labels for complex dataset scenarios, we propose a novel deep learning model named IMAAE (i.e., integrating multiple single-cell datasets via an adversarial autoencoder) to correct the batch effects. After conducting experiments with various dataset scenarios, the results show that IMAAE outperforms existing methods for both qualitative measures and quantitative evaluation. In addition, IMAAE is able to retain both corrected dimension reduction data and corrected gene expression data. These features make it a potential new option for large-scale single-cell gene expression data analysis.

Abstract PD4-06: PD4-06 Obesity-associated changes in transcriptomic profile and immune landscape of primary breast cancer revealed by bulk and single-cell gene expression data

Abstract Background: Breast cancer (BC) is one of the cancer types recognized as an obesity-associated disease. Current understandings of molecular mechanisms underlying the BC-obesity connection however largely came from experimental models while systematic investigation of the impact of obesity on BC biology in large patient series is still lacking. The purpose of this study is to discover changes in the transcriptomic profile of primary BC according to patients’ body mass index (BMI). Data and Methods: Bulk and single-cell gene expression data from treatment-naïve primary breast tumors from non-underweight patients were retrieved from the MINDACT trial (NCT00433589; N = 1481) and the pre-treatment cohort of the BioKey trial (NCT03197389, N = 36), respectively. Three categories were considered for BMI: lean, overweight and obese. The main analyses focused on the invasive carcinoma of no special type (NST) estrogen receptor-positive/HER2-negative (ER+/HER2-, N_bulk = 735, N_single-cell(sc) = 10) and NST ER-/HER2- (N_bulk = 118, N_sc = 15) subgroups. The bulk expression data was subjected to differential gene expression analyses according to BMI which was adjusted for menopausal status and tumor grade, then followed by gene set enrichment analyses. Clustering and cluster annotation were performed on the single-cell profiling data before differentially expressed genes according to BMI were identified for each of the present cell types. Results: Obesity-associated differences in the transcriptomic profile of breast tumors, which were subtle but potentially indicative of a biological relationship, were revealed by the bulk data. In both investigated subgroups, tumors from obese patients were shown to be enriched in cell cycle hallmarks. In ER-/HER2- tumors, adiposity further increased MYC signaling. We also observed different obesity-associated changes according to the ER status. Among ER+/HER2- tumors, those from obese patients were enriched in hallmarks related to inflammatory response compared to those from lean patients. In contrast, these hallmarks appeared to be enriched in the ER-/HER2- tumors from lean patients. Our investigation of the single-cell data further revealed shifts in the cell composition of tumor tissue and cell type-specific transcriptomic differences according to BMI which were more pronounced than those detected from the bulk data. ER+/HER2- tumors from obese patients have a higher frequency of immunosuppressive and pro-tumoral cell subpopulations such as dendritic cells (DC) enriched in immunoregulatory molecules (p = .03), LYVE1+ macrophages (p = .02) and myofibroblasts (p = .03) than those from lean patients. Overexpression of Cyclin D1 and CD24 was found in cancer cells in ER+/HER2- tumors from obese patients. A reduction in anti-tumor immune responses was evident with downregulation of multiple interferons in CD8+ and CD4+ T cells as well as B cells. We observed in the ER-/HER2- subgroup increased infiltration of plasmacytoid DC (p = .01), CCL2+ macrophages (p = .01) in tumors from obese versus lean patients, while fibroblasts showed an opposite tendency. Additionally, significant obesity-associated downregulation of major histocompatibility complex (MHC) molecules class I in cancer cells and MHC class II molecules in B cells could be suggestive of deficient antigen presentation and activation of cytotoxic and helper T cells. Conclusion: We highlighted the impact of obesity on the remodeling of tumor and tumor microenvironment which might generally lead to a suppression of anti-tumor immune responses, albeit potentially via diverse axes according to the ER status. Although investigation on a larger cohort is warranted, our current results suggest that obesity-associated transcriptomic changes in BC could be highly cell type-specific, hence we recommend single-cell approaches in addition to spatial multi-omics analysis to further elucidate the interplay between obesity and BC. Citation Format: Ha-Linh Nguyen, Tatjana Geukens, Marion Maetens, Karen Van Baelen, Maxim De Schepper, Coralie Poncet, Mauro Delorenzi, Marjanka K. Schmid, Emiel Rutgers, Laura Van ’t Veer, Martine Piccart, Fatima Cardoso, Giuseppe Viale, Ayse Bassez, Hanne Vos, Patrick Neven, Ines Nevelsteen, Kevin Punie, Hans Wildiers, Giuseppe Floris, Diether Lambrechts, Ann Smeets, Elia Biganzoli, François Richard, Christine Desmedt. PD4-06 Obesity-associated changes in transcriptomic profile and immune landscape of primary breast cancer revealed by bulk and single-cell gene expression data [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr PD4-06.

Integration of Candida albicans-induced single-cell gene expression data and secretory protein concentrations reveal genetic regulators of inflammation.

Both gene expression and protein concentrations are regulated by genetic variants. Exploring the regulation of both eQTLs and pQTLs simultaneously in a context- and cell-type dependent manner may help to unravel mechanistic basis for genetic regulation of pQTLs. Here, we performed meta-analysis of Candida albicans-induced pQTLs from two population-based cohorts and intersected the results with Candida-induced cell-type specific expression association data (eQTL). This revealed systematic differences between the pQTLs and eQTL, where only 35% of the pQTLs significantly correlated with mRNA expressions at single cell level, indicating the limitation of eQTLs use as a proxy for pQTLs. By taking advantage of the tightly co-regulated pattern of the proteins, we also identified SNPs affecting protein network upon Candida stimulations. Colocalization of pQTLs and eQTLs signals implicated several genomic loci including MMP-1 and AMZ1. Analysis of Candida-induced single cell gene expression data implicated specific cell types that exhibit significant expression QTLs upon stimulation. By highlighting the role of trans-regulatory networks in determining the abundance of secretory proteins, our study serve as a framework to gain insights into the mechanisms of genetic regulation of protein levels in a context-dependent manner.

CCPlotR: an R package for the visualization of cell-cell interactions.

We present CCPlotR-an R package that generates visualizations of cell-cell interactions. CCPlotR is designed to work with the output of tools that predict cell-cell interactions from single-cell gene expression data and requires only a table of predicted interactions as input. The package can generate a comprehensive set of publication-ready figures such as heatmaps, dotplots, circos plots and network diagrams, providing a useful resource for researchers working on cell-cell interactions. CCPlotR is available to download and install from GitHub (https://github.com/Sarah145/CCPlotR) and comes with a toy dataset to demonstrate the different functions. Support for users will be provided via the GitHub issues tracker (https://github.com/Sarah145/CCPlotR/issues).

V-Mapper: topological data analysis for high-dimensional data with velocity

Mapper, a topological data analysis method for high-dimensional data, represents a topological structure as a simplicial complex or graph based on the nerve of clusters. We propose V-Mapper (velocity Mapper), an extension of Mapper, for high-dimensional data with velocity. V-Mapper simultaneously describes a topological structure and flow as a weighted directed graph (V-Mapper graph) by embedding velocity in the edges of the Mapper graph. We apply V-Mapper to single-cell gene expression data using a method for inferring the velocity of gene expression. Moreover, the application of the Hodge decomposition on graph enhances the interpretation of the flow within V-Mapper graph.

Inferring gene regulatory networks from single-cell gene expression data via deep multi-view contrastive learning.

The inference of gene regulatory networks (GRNs) is of great importance for understanding the complex regulatory mechanisms within cells. The emergence of single-cell RNA-sequencing (scRNA-seq) technologies enables the measure of gene expression levels for individual cells, which promotes the reconstruction of GRNs at single-cell resolution. However, existing network inference methods are mainly designed for data collected from a single data source, which ignores the information provided by multiple related data sources. In this paper, we propose a multi-view contrastive learning (DeepMCL) model to infer GRNs from scRNA-seq data collected from multiple data sources or time points. We first represent each gene pair as a set of histogram images, and then introduce a deep Siamese convolutional neural network with contrastive loss to learn the low-dimensional embedding for each gene pair. Moreover, an attention mechanism is introduced to integrate the embeddings extracted from different data sources and different neighbor gene pairs. Experimental results on synthetic and real-world datasets validate the effectiveness of our contrastive learning and attention mechanisms, demonstrating the effectiveness of our model in integrating multiple data sources for GRN inference.

DNA methylation landscape of tumor-associated macrophages reveals pathways, transcription factors and prognostic value relevant to triple-negative breast cancer patients.

The accumulation of myeloid cells, particularly tumor-associated macrophages (TAMs), characterizes the tumor microenvironment (TME) of many solid cancers, including breast cancer. Compared to healthy tissue-resident macrophages, TAMs acquire distinct transcriptomes and tumor-promoting functions by largely unknown mechanisms. Here, we hypothesize the involvement of TME signaling and subsequent epigenetic reprogramming of TAMs. Using the 4T1 mouse model of triple-negative breast cancer, we demonstrate that the presence of cancer cells significantly alters the DNA methylation landscape of macrophages and, to a lesser extent, bone marrow-derived monocytes (BMDMs). TAM methylomes, dissected into BMDM-originating and TAM-specific epigenetic programs, implicated transcription factors (TFs) and signaling pathways involved in TAM reprogramming, correlated with cancer-specific gene expression patterns. Utilizing published single-cell gene expression data, we linked microenvironmentally-derived signals to the cancer-specific DNA methylation landscape of TAMs. These integrative analyses highlighted the role of altered cytokine production in the TME (eg, TGF-β, IFN-γ and CSF1) on the induction of specific TFs (eg, FOSL2, STAT1 and RUNX3) responsible for the epigenetic reprogramming of TAMs. DNA methylation deconvolution identified a TAM-specific signature associated with the identified signaling pathways and TFs, corresponding with severe tumor grade and poor prognosis of breast cancer patients. Similarly, immunosuppressive TAM functions were identified, such as induction of the immune inhibitory receptor-ligand PD-L1 by DNA hypomethylation of Cd274. Collectively, these results provide strong evidence that the epigenetic landscapes of macrophages and monocytes are perturbed by the presence of breast cancer, pointing to molecular mechanisms of TAM reprogramming, impacting patient outcomes.

Single-cell chromatin accessibility identifies enhancer networks driving gene expression during spinal cord development in mouse.

Spinal cord development is precisely orchestrated by spatiotemporal gene regulatory programs. However, the underlying epigenetic mechanisms remain largely elusive. Here, we profiled single-cell chromatin accessibility landscapes in mouse neural tubes spanning embryonic days 9.5-13.5. We identified neuronal-cell-cluster-specific cis-regulatory elements in neural progenitors and neurons. Furthermore, we applied a novel computational method, eNet, to build enhancer networks by integrating single-cell chromatin accessibility and gene expression data and identify the hub enhancers within enhancer networks. It was experimentally validated invivo for Atoh1 that knockout of the hub enhancers, but not the non-hub enhancers, markedly decreased Atoh1 expression and reduced dp1/dI1 cells. Together, our work provides insights into the epigenetic regulation of spinal cord development and a proof-of-concept demonstration of enhancer networks as a general mechanism in transcriptional regulation.

SRGS: sparse partial least squares-based recursive gene selection for gene regulatory network inference

BackgroundThe identification of gene regulatory networks (GRNs) facilitates the understanding of the underlying molecular mechanism of various biological processes and complex diseases. With the availability of single-cell RNA sequencing data, it is essential to infer GRNs from single-cell expression. Although some GRN methods originally developed for bulk expression data can be applicable to single-cell data and several single-cell specific GRN algorithms were developed, recent benchmarking studies have emphasized the need of developing more accurate and robust GRN modeling methods that are compatible for single-cell expression data.ResultsWe present SRGS, SPLS (sparse partial least squares)-based recursive gene selection, to infer GRNs from bulk or single-cell expression data. SRGS recursively selects and scores the genes which may have regulations on the considered target gene based on SPLS. When dealing with gene expression data with dropouts, we randomly scramble samples, set some values in the expression matrix to zeroes, and generate multiple copies of data through multiple iterations to make SRGS more robust. We test SRGS on different kinds of expression data, including simulated bulk data, simulated single-cell data without and with dropouts, and experimental single-cell data, and also compared with the existing GRN methods, including the ones originally developed for bulk data, the ones developed specifically for single-cell data, and even the ones recommended by recent benchmarking studies.ConclusionsIt has been shown that SRGS is competitive with the existing GRN methods and effective in the gene regulatory network inference from bulk or single-cell gene expression data. SRGS is available at: https://github.com/JGuan-lab/SRGS.

Highlights from Recent Cancer Literature

Highlights from Recent Cancer Literature

Single Cell Gene Expression Predicts Plasma Cell Leukemia Patient Response to Daratumumab

The promise of precision medicine is to identify the ideal treatment for each patient, eliminating failed treatment cycles and reducing treatment burdens on patients and payors. While gene sequencing can predict effective treatment options in some cases, studies have estimated that only 10-15% of cancer patients are treated with genotype matched drugs. Gene expression is theoretically an attractive alternative to predicting therapeutic response because, it provides a direct readout of dysregulated cellular activities including druggable pathways. However, the use of bulk gene expression to predict therapeutic response is hampered by tumor heterogeneity; that is, a biopsied tumor contains multiple cells types and different tumor clones and a bulk analysis only generates an average gene expression signal across all of the cells, leading to a decreased sensitivity for identifying biomarkers found only in specific cell subgroups. More recently, single-cell gene expression analysis has shown promise in improving precision medicine through a focused look into the biology of the tumor cells themselves. For example, we previously showed (Scolnick, et al ASCO 2022) that a model trained on single-cell gene expression data collected from 17 Multiple Myeloma (MM) patient's plasma cells (PCs) could be used to predict the therapeutic response of seven independent MM patient samples at 100% accuracy. Here we combine single-cell gene expression analysis with Singleron Biotechnology's MapResponse™ machine learning algorithms to extend our predictions to data from peripheral blood mononuclear cells (PBMCs) of plasma cell leukemia patients who were treated with daratumumab. Our model accurately predicted the response to daratumumab treatment for both patients based on the PCs collected from blood samples. These early findings point to the potential value of combining single-cell RNA sequencing with machine learning methods such as MapResponse™ to bring precision medicine to more patients based on minimal invasive assessment of PBMCs.

EPCO-14. ISOSCELES: AN INTEGRATIVE FRAMEWORK FOR THE PREDICTION OF TREATMENT RESISTANT GLIOBLASTOMA CELLS

Abstract Glioblastoma (GBM) remains the most common and lethal adult primary brain cancer. Two of the most significant issues preventing the development of effective GBM treatments are inter- and intra-tumor heterogeneity. To address these issues, we developed a novel platform termed ISOSCELES (Inferred cell Sensitivity Operating on the integration of Single-Cell Expression and L1000 Expression Signatures). ISOSCELES integrates single-cell gene expression data in individual GBM tumors with perturbation-response data derived from the NIH Library of Integrated Network-Based Cellular Signatures (LINCS) L1000 dataset to predict sensitive and resistant tumor cell populations. Importantly, we analyzed the predictive power of ISOSCELES in an in vivo xenograft model and demonstrated that ISOSCELES reveals the GBM cell identities primed for lineage expansion during treatment with the aurora kinase inhibitor alisertib. These studies suggest that ISOSCELES can be used to identify sensitive and resistant cell populations to targeted therapies in GBM, which can inform treatment decisions in ongoing and future clinical trials.

Pathway trajectory analysis with tensor imputation reveals drug-induced single-cell transcriptomic landscape

Genome-wide identification of single-cell transcriptomic responses of drugs in various human cells is a challenging issue in medical and pharmaceutical research. Here we present a computational method, tensor-based imputation of gene-expression data at the single-cell level (TIGERS), which reveals the drug-induced single-cell transcriptomic landscape. With this algorithm, we predict missing drug-induced single-cell gene-expression data with tensor imputation, and identify trajectories of regulated pathways considering intercellular heterogeneity. Tensor imputation outperformed existing imputation methods for data completion, and provided cell-type-specific transcriptomic responses for unobserved drugs. For example, TIGERS correctly predicted the cell-type-specific expression of maker genes for pancreatic islets. Pathway trajectory analysis of the imputed gene-expression profiles of all combinations of drugs and human cells identified single-cell-specific drug activities and pathway trajectories that reflect drug-induced changes in pathway regulation. The proposed method is expected to expand our understanding of the single-cell mechanisms of drugs at the pathway level.

Time-Varying Gene Expression Network Analysis Reveals Conserved Transition States in Hematopoietic Differentiation between Human and Mouse.

(1) Background: analyses of gene networks can elucidate hematopoietic differentiation from single-cell gene expression data, but most algorithms generate only a single, static network. Because gene interactions change over time, it is biologically meaningful to examine time-varying structures and to capture dynamic, even transient states, and cell-cell relationships. (2) Methods: a transcriptomic atlas of hematopoietic stem and progenitor cells was used for network analysis. After pseudo-time ordering with Monocle 2, LOGGLE was used to infer time-varying networks and to explore changes of differentiation gene networks over time. A range of network analysis tools were used to examine properties and genes in the inferred networks. (3) Results: shared characteristics of attributes during the evolution of differentiation gene networks showed a “U” shape of network density over time for all three branches for human and mouse. Differentiation appeared as a continuous process, originating from stem cells, through a brief transition state marked by fewer gene interactions, before stabilizing in a progenitor state. Human and mouse shared hub genes in evolutionary networks. (4) Conclusions: the conservation of network dynamics in the hematopoietic systems of mouse and human was reflected by shared hub genes and network topological changes during differentiation.

Single-nuclei transcriptome analysis of the shoot apex vascular system differentiation in Populus.

Differentiation of stem cells in the plant apex gives rise to aerial tissues and organs. Presently, we lack a lineage map of the shoot apex cells in woody perennials - a crucial gap considering their role in determining primary and secondary growth. Here, we used single-nuclei RNA-sequencing to determine cell type-specific transcriptomes of the Populus vegetative shoot apex. We identified highly heterogeneous cell populations clustered into seven broad groups represented by 18 transcriptionally distinct cell clusters. Next, we established the developmental trajectories of the epidermis, leaf mesophyll and vascular tissue. Motivated by the high similarities between Populus and Arabidopsis cell population in the vegetative apex, we applied a pipeline for interspecific single-cell gene expression data integration. We contrasted the developmental trajectories of primary phloem and xylem formation in both species, establishing the first comparison of vascular development between a model annual herbaceous and a woody perennial plant species. Our results offer a valuable resource for investigating the principles underlying cell division and differentiation conserved between herbaceous and perennial species while also allowing us to examine species-specific differences at single-cell resolution.

DynDeepDRIM: a dynamic deep learning model to infer direct regulatory interactions using time-course single-cell gene expression data.

Time-course single-cell RNA sequencing (scRNA-seq) data have been widely used to explore dynamic changes in gene expression of transcription factors (TFs) and their target genes. This information is useful to reconstruct cell-type-specific gene regulatory networks (GRNs). However, the existing tools are commonly designed to analyze either time-course bulk gene expression data or static scRNA-seq data via pseudo-time cell ordering. A few methods successfully utilize the information from multiple time points while also considering the characteristics of scRNA-seq data. We proposed dynDeepDRIM, a novel deep learning model to reconstruct GRNs using time-course scRNA-seq data. It represents the joint expression of a gene pair as an image and utilizes the image of the target TF-gene pair and the ones of the potential neighbors to reconstruct GRNs from time-course scRNA-seq data. dynDeepDRIM can effectively remove the transitive TF-gene interactions by considering neighborhood context and model the gene expression dynamics using high-dimensional tensors. We compared dynDeepDRIM with six GRN reconstruction methods on both simulation and four real time-course scRNA-seq data. dynDeepDRIM achieved substantially better performance than the other methods in inferring TF-gene interactions and eliminated the false positives effectively. We also applied dynDeepDRIM to annotate gene functions and found it achieved evidently better performance than the other tools due to considering the neighbor genes.

Identification of spatially variable genes with graph cuts

Single-cell gene expression data with positional information is critical to dissect mechanisms and architectures of multicellular organisms, but the potential is limited by the scalability of current data analysis strategies. Here, we present scGCO, a method based on fast optimization of hidden Markov Random Fields with graph cuts to identify spatially variable genes. Comparing to existing methods, scGCO delivers a superior performance with lower false positive rate and improved specificity, while demonstrates a more robust performance in the presence of noises. Critically, scGCO scales near linearly with inputs and demonstrates orders of magnitude better running time and memory requirement than existing methods, and could represent a valuable solution when spatial transcriptomics data grows into millions of data points and beyond.

SciCAN: single-cell chromatin accessibility and gene expression data integration via cycle-consistent adversarial network

The boom in single-cell technologies has brought a surge of high dimensional data that come from different sources and represent cellular systems from different views. With advances in these single-cell technologies, integrating single-cell data across modalities arises as a new computational challenge. Here, we present an adversarial approach, sciCAN, to integrate single-cell chromatin accessibility and gene expression data in an unsupervised manner. We benchmarked sciCAN with 5 existing methods in 5 scATAC-seq/scRNA-seq datasets, and we demonstrated that our method dealt with data integration with consistent performance across datasets and better balance of mutual transferring between modalities than the other 5 existing methods. We further applied sciCAN to 10X Multiome data and confirmed that the integrated representation preserves biological relationships within the hematopoietic hierarchy. Finally, we investigated CRISPR-perturbed single-cell K562 ATAC-seq and RNA-seq data to identify cells with related responses to different perturbations in these different modalities.