Probabilistic embedding, clustering, and alignment for integrating spatial transcriptomics data with PRECAST

Computational modeling for deciphering tissue microenvironment heterogeneity from spatially resolved transcriptomics

Spatial transcriptomics techniques, while measuring gene expression, retain spatial location information, aiding in situ studies of organismal tissue architecture and the progression of pathological processes. These techniques generate vast amounts of omics data, necessitating the development of computational methods to reveal the underlying tissue microenvironment heterogeneity. The main directions in spatial transcriptomics data analysis are spatial domain detection and spatial deconvolution, which can identify spatial functional regions and parse the distribution of cell types in spatial transcriptomics data by integrating single-cell transcriptomics data. In these two research directions, many computational methods have been successively proposed. This article will categorize them into three types: machine learning-based methods, probabilistic models-based methods, and deep learning-based methods. It will list and discuss the representative algorithms of each type along with their advantages and disadvantages and describe the datasets and evaluation metrics used to assess these computational methods, facilitating researchers in selecting suitable computational methods according to their research needs. Finally, combining the latest technological developments and the advantages and disadvantages of current algorithms, this article will look forward to the future directions of computational method development.

Computational methods for alignment and integration of spatially resolved transcriptomics data

Most of the complex biological regulatory activities occur in three dimensions (3D). To better analyze biological processes, it is essential not only to decipher the molecular information of numerous cells but also to understand how their spatial contexts influence their behavior. With the development of spatially resolved transcriptomics (SRT) technologies, SRT datasets are being generated to simultaneously characterize gene expression and spatial arrangement information within tissues, organs or organisms. To fully leverage spatial information, the focus extends beyond individual two-dimensional (2D) slices. Two tasks known as slices alignment and data integration have been introduced to establish correlations between multiple slices, enhancing the effectiveness of downstream tasks. Currently, numerous related methods have been developed. In this review, we first elucidate the details and principles behind several representative methods. Then we report the testing results of these methods on various SRT datasets, and assess their performance in representative downstream tasks. Insights into the strengths and weaknesses of each method and the reasons behind their performance are discussed. Finally, we provide an outlook on future developments. The codes and details of experiments are now publicly available at https://github.com/YangLabHKUST/SRT_alignment_and_integration.

High-Dimensional Overdispersed Generalized Factor Model With Application to Single-Cell Sequencing Data Analysis

The current high-dimensional linear factor models fail to account for the different types of variables, while high-dimensional nonlinear factor models often overlook the overdispersion present in mixed-type data. However, overdispersion is prevalent in practical applications, particularly in fields like biomedical and genomics studies. To address this practical demand, we propose an overdispersed generalized factor model (OverGFM) for performing high-dimensional nonlinear factor analysis on overdispersed mixed-type data. Our approach incorporates an additional error term to capture the overdispersion that cannot be accounted for by factors alone. However, this introduces significant computational challenges due to the involvement of two high-dimensional latent random matrices in the nonlinear model. To overcome these challenges, we propose a novel variational EM algorithm that integrates Laplace and Taylor approximations. This algorithm provides iterative explicit solutions for the complex variational parameters and is proven to possess excellent convergence properties. We also develop a criterion based on the singular value ratio to determine the optimal number of factors. Numerical results demonstrate the effectiveness of this criterion. Through comprehensive simulation studies, we show that OverGFM outperforms state-of-the-art methods in terms of estimation accuracy and computational efficiency. Furthermore, we demonstrate the practical merit of our method through its application to two datasets from genomics. To facilitate its usage, we have integrated the implementation of OverGFM into the R package GFM.

Statistical batch-aware embedded integration, dimension reduction, and alignment for spatial transcriptomics

MOTIVATION

Spatial transcriptomics (ST) technologies provide richer insights into the molecular characteristics of cells by simultaneously measuring gene expression profiles and their relative locations. However, each slice can only contain limited biological variation, and since there are almost always non-negligible batch effects across different slices, integrating numerous slices to account for batch effects and locations is not straightforward. Performing multi-slice integration, dimensionality reduction, and other downstream analyses separately often results in suboptimal embeddings for technical artifacts and biological variations. Joint modeling integrating these steps can enhance our understanding of the complex interplay between technical artifacts and biological signals, leading to more accurate and insightful results.

RESULTS

In this context, we propose a hierarchical hidden Markov random field model STADIA to reduce batch effects, extract common biological patterns across multiple ST slices, and simultaneously identify spatial domains. We demonstrate the effectiveness of STADIA using five datasets from different species (human and mouse), various organs (brain, skin, and liver), and diverse platforms (10x Visium, ST, and Slice-seqV2). STADIA can capture common tissue structures across multiple slices and preserve slice-specific biological signals. In addition, STADIA outperforms the other three competing methods (PRECAST, fastMNN, and Harmony) in terms of the balance between batch mixing and spatial domain identification, and it demonstrates the advantage of joint modeling when compared to STAGATE and GraphST.

AVAILABILITY AND IMPLEMENTATION

The source code implemented by R is available at https://github.com/zhanglabtools/STADIA and archived with version 1.01 on Zenodo https://zenodo.org/records/13637744.

INSPIRE: interpretable, flexible and spatially-aware integration of multiple spatial transcriptomics datasets from diverse sources

Recent advances in spatial transcriptomics technologies have led to a growing number of diverse datasets, offering unprecedented opportunities to explore tissue organizations and functions within spatial contexts. However, it remains a significant challenge to effectively integrate and interpret these data, often originating from different samples, technologies, and developmental stages. In this paper, we present INSPIRE, a deep learning method for integrative analyses of multiple spatial transcriptomics datasets to address this challenge. With designs of graph neural networks and an adversarial learning mechanism, INSPIRE enables spatially informed and adaptable integration of data from varying sources. By incorporating non-negative matrix factorization, INSPIRE uncovers interpretable spatial factors with corresponding gene programs, revealing tissue architectures, cell type distributions and biological processes. We demonstrate the capabilities of INSPIRE by applying it to human cortex slices from different samples, mouse brain slices with complementary views, mouse hippocampus and embryo slices generated through different technologies, and spatiotemporal organogenesis atlases containing half a million spatial spots. INSPIRE shows superior performance in identifying detailed biological signals, effectively borrowing information across distinct profiling technologies, and elucidating dynamical changes during embryonic development. Furthermore, we utilize INSPIRE to build 3D models of tissues and whole organisms from multiple slices, demonstrating its power and versatility.

A graph self-supervised residual learning framework for domain identification and data integration of spatial transcriptomics

Spatial transcriptomics (ST) technologies allow for comprehensive characterization of gene expression patterns in the context of tissue microenvironment. However, accurately identifying domains with spatial coherence in both gene expression and histology in situ and effectively integrating data from multi-sample remains challenging. Here, we propose ResST, a graph self-supervised residual learning model based on graph neural network and Margin Disparity Discrepancy (MDD) theory. ResST aggregates gene expression, biological effects, spatial location, and morphological information to capture nonlinear relationships between a cell and surrounding cells for spatial domain identification. Also, ResST integrates multiple ST datasets and aligns latent embeddings based on MDD theory for correcting batch effects. Results show that ResST identifies continuous spatial domains at a finer scale in ten ST datasets acquired with different technologies. Moreover, ResST efficiently integrated data from multiple tissue sections vertically or horizontally while correcting batch effects. Overall, ResST demonstrates exceptional performance in analyzing ST datasets.

Systematic comparison of sequencing-based spatial transcriptomic methods

Recent developments of sequencing-based spatial transcriptomics (sST) have catalyzed important advancements by facilitating transcriptome-scale spatial gene expression measurement. Despite this progress, efforts to comprehensively benchmark different platforms are currently lacking. The extant variability across technologies and datasets poses challenges in formulating standardized evaluation metrics. In this study, we established a collection of reference tissues and regions characterized by well-defined histological architectures, and used them to generate data to compare 11 sST methods. We highlighted molecular diffusion as a variable parameter across different methods and tissues, significantly affecting the effective resolutions. Furthermore, we observed that spatial transcriptomic data demonstrate unique attributes beyond merely adding a spatial axis to single-cell data, including an enhanced ability to capture patterned rare cell states along with specific markers, albeit being influenced by multiple factors including sequencing depth and resolution. Our study assists biologists in sST platform selection, and helps foster a consensus on evaluation standards and establish a framework for future benchmarking efforts that can be used as a gold standard for the development and benchmarking of computational tools for spatial transcriptomic analysis.

Benchmarking clustering, alignment, and integration methods for spatial transcriptomics

BACKGROUND

Spatial transcriptomics (ST) is advancing our understanding of complex tissues and organisms. However, building a robust clustering algorithm to define spatially coherent regions in a single tissue slice and aligning or integrating multiple tissue slices originating from diverse sources for essential downstream analyses remains challenging. Numerous clustering, alignment, and integration methods have been specifically designed for ST data by leveraging its spatial information. The absence of comprehensive benchmark studies complicates the selection of methods and future method development.

RESULTS

In this study, we systematically benchmark a variety of state-of-the-art algorithms with a wide range of real and simulated datasets of varying sizes, technologies, species, and complexity. We analyze the strengths and weaknesses of each method using diverse quantitative and qualitative metrics and analyses, including eight metrics for spatial clustering accuracy and contiguity, uniform manifold approximation and projection visualization, layer-wise and spot-to-spot alignment accuracy, and 3D reconstruction, which are designed to assess method performance as well as data quality. The code used for evaluation is available on our GitHub. Additionally, we provide online notebook tutorials and documentation to facilitate the reproduction of all benchmarking results and to support the study of new methods and new datasets.

CONCLUSIONS

Our analyses lead to comprehensive recommendations that cover multiple aspects, helping users to select optimal tools for their specific needs and guide future method development.

Sensitive Multiplexed MicroRNA Spatial Profiling and Data Classification Framework Applied to Murine Breast Tumors

MicroRNAs (miRNAs) are small RNAs that are often dysregulated in many diseases, including cancers. They are highly tissue-specific and stable, thus, making them particularly useful as biomarkers. As the spatial transcriptomics field advances, protocols that enable highly sensitive and spatially resolved detection become necessary to maximize the information gained from samples. This is especially true of miRNAs where the location their expression within tissue can provide prognostic value with regard to patient outcome. Equally as important as detection are ways to assess and visualize the miRNA's spatial information in order to leverage the power of spatial transcriptomics over that of traditional nonspatial bulk assays. We present a highly sensitive methodology that simultaneously quantitates and spatially detects seven miRNAs in situ on formalin-fixed paraffin-embedded tissue sections. This method utilizes rolling circle amplification (RCA) in conjunction with a dual scanning approach in nanoliter well arrays with embedded hydrogel posts. The hydrogel posts are functionalized with DNA probes that enable the detection of miRNAs across a large dynamic range (4 orders of magnitude) and a limit of detection of 0.17 zeptomoles (1.7 × 10-4 attomoles). We applied our methodology coupled with a data analysis pipeline to K14-Cre Brca1f/fTp53f/f murine breast tumors to showcase the information gained from this approach.

Spatially Resolved Single-Cell Omics: Methods, Challenges, and Future Perspectives

Overlaying omics data onto spatial biological dimensions has been a promising technology to provide high-resolution insights into the interactome and cellular heterogeneity relative to the organization of the molecular microenvironment of tissue samples in normal and disease states. Spatial omics can be categorized into three major modalities: (a) next-generation sequencing-based assays, (b) imaging-based spatially resolved transcriptomics approaches including in situ hybridization/in situ sequencing, and (c) imaging-based spatial proteomics. These modalities allow assessment of transcripts and proteins at a cellular level, generating large and computationally challenging datasets. The lack of standardized computational pipelines to analyze and integrate these nonuniform structured data has made it necessary to apply artificial intelligence and machine learning strategies to best visualize and translate their complexity. In this review, we summarize the currently available techniques and computational strategies, highlight their advantages and limitations, and discuss their future prospects in the scientific field.

Integrating spatially-resolved transcriptomics data across tissues and individuals: challenges and opportunities

Advances in spatially-resolved transcriptomics (SRT) technologies have propelled the development of new computational analysis methods to unlock biological insights. As the cost of generating these data decreases, these technologies provide an exciting opportunity to create large-scale atlases that integrate SRT data across multiple tissues, individuals, species, or phenotypes to perform population-level analyses. Here, we describe unique challenges of varying spatial resolutions in SRT data, as well as highlight the opportunities for standardized preprocessing methods along with computational algorithms amenable to atlas-scale datasets leading to improved sensitivity and reproducibility in the future.

Advancing immunotherapy for melanoma: the critical role of single-cell analysis in identifying predictive biomarkers

Melanoma, a malignant skin cancer arising from melanocytes, exhibits rapid metastasis and a high mortality rate, especially in advanced stages. Current treatment modalities, including surgery, radiation, and immunotherapy, offer limited success, with immunotherapy using immune checkpoint inhibitors (ICIs) being the most promising. However, the high mortality rate underscores the urgent need for robust, non-invasive biomarkers to predict patient response to adjuvant therapies. The immune microenvironment of melanoma comprises various immune cells, which influence tumor growth and immune response. Melanoma cells employ multiple mechanisms for immune escape, including defects in immune recognition and epithelial-mesenchymal transition (EMT), which collectively impact treatment efficacy. Single-cell analysis technologies, such as single-cell RNA sequencing (scRNA-seq), have revolutionized the understanding of tumor heterogeneity and immune microenvironment dynamics. These technologies facilitate the identification of rare cell populations, co-expression patterns, and regulatory networks, offering deep insights into tumor progression, immune response, and therapy resistance. In the realm of biomarker discovery for melanoma, single-cell analysis has demonstrated significant potential. It aids in uncovering cellular composition, gene profiles, and novel markers, thus advancing diagnosis, treatment, and prognosis. Additionally, tumor-associated antibodies and specific genetic and cellular markers identified through single-cell analysis hold promise as predictive biomarkers. Despite these advancements, challenges such as RNA-protein expression discrepancies and tumor heterogeneity persist, necessitating further research. Nonetheless, single-cell analysis remains a powerful tool in elucidating the mechanisms underlying therapy response and resistance, ultimately contributing to the development of personalized melanoma therapies and improved patient outcomes.

Beaconet: A Reference-Free Method for Integrating Multiple Batches of Single-Cell Transcriptomic Data in Original Molecular Space

Integrating multiple single-cell datasets is essential for the comprehensive understanding of cell heterogeneity. Batch effect is the undesired systematic variations among technologies or experimental laboratories that distort biological signals and hinder the integration of single-cell datasets. However, existing methods typically rely on a selected dataset as a reference, leading to inconsistent integration performance using different references, or embed cells into uninterpretable low-dimensional feature space. To overcome these limitations, a reference-free method, Beaconet, for integrating multiple single-cell transcriptomic datasets in original molecular space by aligning the global distribution of each batch using an adversarial correction network is presented. Through extensive comparisons with 13 state-of-the-art methods, it is demonstrated that Beaconet can effectively remove batch effect while preserving biological variations and is superior to existing unsupervised methods using all possible references in overall performance. Furthermore, Beaconet performs integration in the original molecular feature space, enabling the characterization of cell types and downstream differential expression analysis directly using integrated data with gene-expression features. Additionally, when applying to large-scale atlas data integration, Beaconet shows notable advantages in both time- and space-efficiencies. In summary, Beaconet serves as an effective and efficient batch effect removal tool that can facilitate the integration of single-cell datasets in a reference-free and molecular feature-preserved mode.

Simulating multiple variability in spatially resolved transcriptomics with scCube

A pressing challenge in spatially resolved transcriptomics (SRT) is to benchmark the computational methods. A widely-used approach involves utilizing simulated data. However, biases exist in terms of the currently available simulated SRT data, which seriously affects the accuracy of method evaluation and validation. Herein, we present scCube ( https://github.com/ZJUFanLab/scCube ), a Python package for independent, reproducible, and technology-diverse simulation of SRT data. scCube not only enables the preservation of spatial expression patterns of genes in reference-based simulations, but also generates simulated data with different spatial variability (covering the spatial pattern type, the resolution, the spot arrangement, the targeted gene type, and the tissue slice dimension, etc.) in reference-free simulations. We comprehensively benchmark scCube with existing single-cell or SRT simulators, and demonstrate the utility of scCube in benchmarking spot deconvolution, gene imputation, and resolution enhancement methods in detail through three applications.

A data-driven single-cell and spatial transcriptomic map of the human prefrontal cortex

The molecular organization of the human neocortex historically has been studied in the context of its histological layers. However, emerging spatial transcriptomic technologies have enabled unbiased identification of transcriptionally defined spatial domains that move beyond classic cytoarchitecture. We used the Visium spatial gene expression platform to generate a data-driven molecular neuroanatomical atlas across the anterior-posterior axis of the human dorsolateral prefrontal cortex. Integration with paired single-nucleus RNA-sequencing data revealed distinct cell type compositions and cell-cell interactions across spatial domains. Using PsychENCODE and publicly available data, we mapped the enrichment of cell types and genes associated with neuropsychiatric disorders to discrete spatial domains.

NetActivity enhances transcriptional signals by combining gene expression into robust gene set activity scores through interpretable autoencoders

Grouping gene expression into gene set activity scores (GSAS) provides better biological insights than studying individual genes. However, existing gene set projection methods cannot return representative, robust, and interpretable GSAS. We developed NetActivity, a machine learning framework that generates GSAS based on a sparsely-connected autoencoder, where each neuron in the inner layer represents a gene set. We proposed a three-tier training that yielded representative, robust, and interpretable GSAS. NetActivity model was trained with 1518 GO biological processes terms and KEGG pathways and all GTEx samples. NetActivity generates GSAS robust to the initialization parameters and representative of the original transcriptome, and assigned higher importance to more biologically relevant genes. Moreover, NetActivity returns GSAS with a more consistent definition and higher interpretability than GSVA and hipathia, state-of-the-art gene set projection methods. Finally, NetActivity enables combining bulk RNA-seq and microarray datasets in a meta-analysis of prostate cancer progression, highlighting gene sets related to cell division, key for disease progression. When applied to metastatic prostate cancer, gene sets associated with cancer progression were also altered due to drug resistance, while a classical enrichment analysis identified gene sets irrelevant to the phenotype. NetActivity is publicly available in Bioconductor and GitHub.

An integrated single-nucleus and spatial transcriptomics atlas reveals the molecular landscape of the human hippocampus

The hippocampus contains many unique cell types, which serve the structure's specialized functions, including learning, memory and cognition. These cells have distinct spatial topography, morphology, physiology, and connectivity, highlighting the need for transcriptome-wide profiling strategies that retain cytoarchitectural organization. Here, we generated spatially-resolved transcriptomics (SRT) and single-nucleus RNA-sequencing (snRNA-seq) data from adjacent tissue sections of the anterior human hippocampus across ten adult neurotypical donors. We defined molecular profiles for hippocampal cell types and spatial domains. Using non-negative matrix factorization and transfer learning, we integrated these data to define gene expression patterns within the snRNA-seq data and infer the expression of these patterns in the SRT data. With this approach, we leveraged existing rodent datasets that feature information on circuit connectivity and neural activity induction to make predictions about axonal projection targets and likelihood of ensemble recruitment in spatially-defined cellular populations of the human hippocampus. Finally, we integrated genome-wide association studies with transcriptomic data to identify enrichment of genetic components for neurodevelopmental, neuropsychiatric, and neurodegenerative disorders across cell types, spatial domains, and gene expression patterns of the human hippocampus. To make this comprehensive molecular atlas accessible to the scientific community, both raw and processed data are freely available, including through interactive web applications.

Identification of spatially-resolved markers of malignant transformation in Intraductal Papillary Mucinous Neoplasms

The existing Intraductal Papillary Mucinous Neoplasm (IPMN) risk stratification relies on clinical and histological factors, resulting in inaccuracies and leading to suboptimal treatment. This is due to the lack of appropriate molecular markers that can guide patients toward the best therapeutic options. Here, we assess and confirm subtype-specific markers for IPMN across two independent cohorts of patients using two Spatial Transcriptomics (ST) technologies. Specifically, we identify HOXB3 and ZNF117 as markers for Low-Grade Dysplasia, SPDEF and gastric neck cell markers in borderline cases, and NKX6-2 and gastric isthmus cell markers in High-Grade-Dysplasia Gastric IPMN, highlighting the role of TNFα and MYC activation in IPMN progression and the role of NKX6-2 in the specific Gastric IPMN progression. In conclusion, our work provides a step forward in understanding the gene expression landscapes of IPMN and the critical transcriptional networks related to PDAC progression.

High-dimensional covariate-augmented overdispersed poisson factor model

The current Poisson factor models often assume that the factors are unknown, which overlooks the explanatory potential of certain observable covariates. This study focuses on high dimensional settings, where the number of the count response variables and/or covariates can diverge as the sample size increases. A covariate-augmented overdispersed Poisson factor model is proposed to jointly perform a high-dimensional Poisson factor analysis and estimate a large coefficient matrix for overdispersed count data. A group of identifiability conditions is provided to theoretically guarantee computational identifiability. We incorporate the interdependence of both response variables and covariates by imposing a low-rank constraint on the large coefficient matrix. To address the computation challenges posed by nonlinearity, two high-dimensional latent matrices, and the low-rank constraint, we propose a novel variational estimation scheme that combines Laplace and Taylor approximations. We also develop a criterion based on a singular value ratio to determine the number of factors and the rank of the coefficient matrix. Comprehensive simulation studies demonstrate that the proposed method outperforms the state-of-the-art methods in estimation accuracy and computational efficiency. The practical merit of our method is demonstrated by an application to the CITE-seq dataset. A flexible implementation of our proposed method is available in the R package COAP.

Spatial transcriptomics data and analytical methods: An updated perspective

Spatial transcriptomics (ST) is a newly emerging field that integrates high-resolution imaging and transcriptomic data to enable the high-throughput analysis of the spatial localization of transcripts in diverse biological systems. The rapid progress in this field necessitates the development of innovative computational methods to effectively tackle the distinct challenges posed by the analysis of ST data. These platforms, integrating AI techniques, offer a promising avenue for understanding disease mechanisms and expediting drug discovery. Despite significant advances in the development of ST data analysis techniques, there is an ongoing need to enhance these models for increased biological relevance. In this review, we briefly discuss the ST-related databases and current deep-learning-based models for spatial transcriptome data analyses and highlight their roles and future perspectives in biomedical applications.

Navigating the landscapes of spatial transcriptomics: How computational methods guide the way

Spatially resolved transcriptomics has been dramatically transforming biological and medical research in various fields. It enables transcriptome profiling at single-cell, multi-cellular, or sub-cellular resolution, while retaining the information of geometric localizations of cells in complex tissues. The coupling of cell spatial information and its molecular characteristics generates a novel multi-modal high-throughput data source, which poses new challenges for the development of analytical methods for data-mining. Spatial transcriptomic data are often highly complex, noisy, and biased, presenting a series of difficulties, many unresolved, for data analysis and generation of biological insights. In addition, to keep pace with the ever-evolving spatial transcriptomic experimental technologies, the existing analytical theories and tools need to be updated and reformed accordingly. In this review, we provide an overview and discussion of the current computational approaches for mining of spatial transcriptomics data. Future directions and perspectives of methodology design are proposed to stimulate further discussions and advances in new analytical models and algorithms. This article is categorized under: RNA Methods > RNA Analyses in Cells RNA Evolution and Genomics > Computational Analyses of RNA RNA Export and Localization > RNA Localization.

BatchEval Pipeline: batch effect evaluation workflow for multiple datasets joint analysis

As genomic sequencing technology continues to advance, it becomes increasingly important to perform joint analyses of multiple datasets of transcriptomics. However, batch effect presents challenges for dataset integration, such as sequencing data measured on different platforms, and datasets collected at different times. Here, we report the development of BatchEval Pipeline, a batch effect workflow used to evaluate batch effect on dataset integration. The BatchEval Pipeline generates a comprehensive report, which consists of a series of HTML pages for assessment findings, including a main page, a raw dataset evaluation page, and several built-in methods evaluation pages. The main page exhibits basic information of the integrated datasets, a comprehensive score of batch effect, and the most recommended method for removing batch effect from the current datasets. The remaining pages exhibit evaluation details for the raw dataset, and evaluation results from the built-in batch effect removal methods after removing batch effect. This comprehensive report enables researchers to accurately identify and remove batch effects, resulting in more reliable and meaningful biological insights from integrated datasets. In summary, the BatchEval Pipeline represents a significant advancement in batch effect evaluation, and is a valuable tool to improve the accuracy and reliability of the experimental results.

Availability & Implementation

The source code of the BatchEval Pipeline is available at https://github.com/STOmics/BatchEval.

DESpace: spatially variable gene detection via differential expression testing of spatial clusters

MOTIVATION

Spatially resolved transcriptomics (SRT) enables scientists to investigate spatial context of mRNA abundance, including identifying spatially variable genes (SVGs), i.e. genes whose expression varies across the tissue. Although several methods have been proposed for this task, native SVG tools cannot jointly model biological replicates, or identify the key areas of the tissue affected by spatial variability.

RESULTS

Here, we introduce DESpace, a framework, based on an original application of existing methods, to discover SVGs. In particular, our approach inputs all types of SRT data, summarizes spatial information via spatial clusters, and identifies spatially variable genes by performing differential gene expression testing between clusters. Furthermore, our framework can identify (and test) the main cluster of the tissue affected by spatial variability; this allows scientists to investigate spatial expression changes in specific areas of interest. Additionally, DESpace enables joint modeling of multiple samples (i.e. biological replicates); compared to inference based on individual samples, this approach increases statistical power, and targets SVGs with consistent spatial patterns across replicates. Overall, in our benchmarks, DESpace displays good true positive rates, controls for false positive and false discovery rates, and is computationally efficient.

AVAILABILITY AND IMPLEMENTATION

DESpace is freely distributed as a Bioconductor R package at https://bioconductor.org/packages/DESpace.

spatiAlign: an unsupervised contrastive learning model for data integration of spatially resolved transcriptomics

BACKGROUND

Integrative analysis of spatially resolved transcriptomics datasets empowers a deeper understanding of complex biological systems. However, integrating multiple tissue sections presents challenges for batch effect removal, particularly when the sections are measured by various technologies or collected at different times.

FINDINGS

We propose spatiAlign, an unsupervised contrastive learning model that employs the expression of all measured genes and the spatial location of cells, to integrate multiple tissue sections. It enables the joint downstream analysis of multiple datasets not only in low-dimensional embeddings but also in the reconstructed full expression space.

CONCLUSIONS

In benchmarking analysis, spatiAlign outperforms state-of-the-art methods in learning joint and discriminative representations for tissue sections, each potentially characterized by complex batch effects or distinct biological characteristics. Furthermore, we demonstrate the benefits of spatiAlign for the integrative analysis of time-series brain sections, including spatial clustering, differential expression analysis, and particularly trajectory inference that requires a corrected gene expression matrix.

Challenges and opportunities to computationally deconvolve heterogeneous tissue with varying cell sizes using single-cell RNA-sequencing datasets

Deconvolution of cell mixtures in "bulk" transcriptomic samples from homogenate human tissue is important for understanding disease pathologies. However, several experimental and computational challenges impede transcriptomics-based deconvolution approaches using single-cell/nucleus RNA-seq reference atlases. Cells from the brain and blood have substantially different sizes, total mRNA, and transcriptional activities, and existing approaches may quantify total mRNA instead of cell type proportions. Further, standards are lacking for the use of cell reference atlases and integrative analyses of single-cell and spatial transcriptomics data. We discuss how to approach these key challenges with orthogonal "gold standard" datasets for evaluating deconvolution methods.

Probabilistic cell/domain-type assignment of spatial transcriptomics data with SpatialAnno

In the analysis of both single-cell RNA sequencing (scRNA-seq) and spatially resolved transcriptomics (SRT) data, classifying cells/spots into cell/domain types is an essential analytic step for many secondary analyses. Most of the existing annotation methods have been developed for scRNA-seq datasets without any consideration of spatial information. Here, we present SpatialAnno, an efficient and accurate annotation method for spatial transcriptomics datasets, with the capability to effectively leverage a large number of non-marker genes as well as 'qualitative' information about marker genes without using a reference dataset. Uniquely, SpatialAnno estimates low-dimensional embeddings for a large number of non-marker genes via a factor model while promoting spatial smoothness among neighboring spots via a Potts model. Using both simulated and four real spatial transcriptomics datasets from the 10x Visium, ST, Slide-seqV1/2, and seqFISH platforms, we showcase the method's improved spatial annotation accuracy, including its robustness to the inclusion of marker genes for irrelevant cell/domain types and to various degrees of marker gene misspecification. SpatialAnno is computationally scalable and applicable to SRT datasets from different platforms. Furthermore, the estimated embeddings for cellular biological effects facilitate many downstream analyses.

SPACEL: deep learning-based characterization of spatial transcriptome architectures

Spatial transcriptomics (ST) technologies detect mRNA expression in single cells/spots while preserving their two-dimensional (2D) spatial coordinates, allowing researchers to study the spatial distribution of the transcriptome in tissues; however, joint analysis of multiple ST slices and aligning them to construct a three-dimensional (3D) stack of the tissue still remain a challenge. Here, we introduce spatial architecture characterization by deep learning (SPACEL) for ST data analysis. SPACEL comprises three modules: Spoint embeds a multiple-layer perceptron with a probabilistic model to deconvolute cell type composition for each spot in a single ST slice; Splane employs a graph convolutional network approach and an adversarial learning algorithm to identify spatial domains that are transcriptomically and spatially coherent across multiple ST slices; and Scube automatically transforms the spatial coordinate systems of consecutive slices and stacks them together to construct a 3D architecture of the tissue. Comparisons against 19 state-of-the-art methods using both simulated and real ST datasets from various tissues and ST technologies demonstrate that SPACEL outperforms the others for cell type deconvolution, for spatial domain identification, and for 3D alignment, thus showcasing SPACEL as a valuable integrated toolkit for ST data processing and analysis.

Mapping the topography of spatial gene expression with interpretable deep learning

Spatially resolved transcriptomics technologies provide high-throughput measurements of gene expression in a tissue slice, but the sparsity of this data complicates the analysis of spatial gene expression patterns such as gene expression gradients. We address these issues by deriving a topographic map of a tissue slice-analogous to a map of elevation in a landscape-using a novel quantity called the isodepth. Contours of constant isodepth enclose spatial domains with distinct cell type composition, while gradients of the isodepth indicate spatial directions of maximum change in gene expression. We develop GASTON, an unsupervised and interpretable deep learning algorithm that simultaneously learns the isodepth, spatial gene expression gradients, and piecewise linear functions of the isodepth that model both continuous gradients and discontinuous spatial variation in the expression of individual genes. We validate GASTON by showing that it accurately identifies spatial domains and marker genes across several biological systems. In SRT data from the brain, GASTON reveals gradients of neuronal differentiation and firing, and in SRT data from a tumor sample, GASTON infers gradients of metabolic activity and epithelial-mesenchymal transition (EMT)-related gene expression in the tumor microenvironment.

Challenges and opportunities to computationally deconvolve heterogeneous tissue with varying cell sizes using single cell RNA-sequencing datasets

Deconvolution of cell mixtures in "bulk" transcriptomic samples from homogenate human tissue is important for understanding the pathologies of diseases. However, several experimental and computational challenges remain in developing and implementing transcriptomics-based deconvolution approaches, especially those using a single cell/nuclei RNA-seq reference atlas, which are becoming rapidly available across many tissues. Notably, deconvolution algorithms are frequently developed using samples from tissues with similar cell sizes. However, brain tissue or immune cell populations have cell types with substantially different cell sizes, total mRNA expression, and transcriptional activity. When existing deconvolution approaches are applied to these tissues, these systematic differences in cell sizes and transcriptomic activity confound accurate cell proportion estimates and instead may quantify total mRNA content. Furthermore, there is a lack of standard reference atlases and computational approaches to facilitate integrative analyses, including not only bulk and single cell/nuclei RNA-seq data, but also new data modalities from spatial -omic or imaging approaches. New multi-assay datasets need to be collected with orthogonal data types generated from the same tissue block and the same individual, to serve as a "gold standard" for evaluating new and existing deconvolution methods. Below, we discuss these key challenges and how they can be addressed with the acquisition of new datasets and approaches to analysis.

The promise and challenge of spatial omics in dissecting tumour microenvironment and the role of AI

Growing evidence supports the critical role of tumour microenvironment (TME) in tumour progression, metastases, and treatment response. However, the in-situ interplay among various TME components, particularly between immune and tumour cells, are largely unknown, hindering our understanding of how tumour progresses and responds to treatment. While mainstream single-cell omics techniques allow deep, single-cell phenotyping, they lack crucial spatial information for in-situ cell-cell interaction analysis. On the other hand, tissue-based approaches such as hematoxylin and eosin and chromogenic immunohistochemistry staining can preserve the spatial information of TME components but are limited by their low-content staining. High-content spatial profiling technologies, termed spatial omics, have greatly advanced in the past decades to overcome these limitations. These technologies continue to emerge to include more molecular features (RNAs and/or proteins) and to enhance spatial resolution, opening new opportunities for discovering novel biological knowledge, biomarkers, and therapeutic targets. These advancements also spur the need for novel computational methods to mine useful TME insights from the increasing data complexity confounded by high molecular features and spatial resolution. In this review, we present state-of-the-art spatial omics technologies, their applications, major strengths, and limitations as well as the role of artificial intelligence (AI) in TME studies.

Integrated single cell and unsupervised spatial transcriptomic analysis defines molecular anatomy of the human dorsolateral prefrontal cortex

Generation of a molecular neuroanatomical map of the human prefrontal cortex reveals novel spatial domains and cell-cell interactions relevant for psychiatric disease. The molecular organization of the human neocortex has been historically studied in the context of its histological layers. However, emerging spatial transcriptomic technologies have enabled unbiased identification of transcriptionally-defined spatial domains that move beyond classic cytoarchitecture. Here we used the Visium spatial gene expression platform to generate a data-driven molecular neuroanatomical atlas across the anterior-posterior axis of the human dorsolateral prefrontal cortex (DLPFC). Integration with paired single nucleus RNA-sequencing data revealed distinct cell type compositions and cell-cell interactions across spatial domains. Using PsychENCODE and publicly available data, we map the enrichment of cell types and genes associated with neuropsychiatric disorders to discrete spatial domains. Finally, we provide resources for the scientific community to explore these integrated spatial and single cell datasets at research.libd.org/spatialDLPFC/.