Advances in spatial transcriptomics and related data analysis strategies
Journal of Translational Medicine volume 21, Article number: 330 (2023)
Spatial transcriptomics technologies developed in recent years can provide various information including tissue heterogeneity, which is fundamental in biological and medical research, and have been making significant breakthroughs. Single-cell RNA sequencing (scRNA-seq) cannot provide spatial information, while spatial transcriptomics technologies allow gene expression information to be obtained from intact tissue sections in the original physiological context at a spatial resolution. Various biological insights can be generated into tissue architecture and further the elucidation of the interaction between cells and the microenvironment. Thus, we can gain a general understanding of histogenesis processes and disease pathogenesis, etc. Furthermore, in silico methods involving the widely distributed R and Python packages for data analysis play essential roles in deriving indispensable bioinformation and eliminating technological limitations. In this review, we summarize available technologies of spatial transcriptomics, probe into several applications, discuss the computational strategies and raise future perspectives, highlighting the developmental potential.
Human organs and systems are comprised of distinct cell subpopulations whose physiological processes and functions are deeply correlated with their spatial distributions and cellular interactions. To gain a deeper understanding of tissue architecture as well as heterogeneity and to subsequently obtain biological insights into intercellular communication and microenvironment, it is crucial to decipher the disparities among tissue regions and cells in their original spatial context. Previously developed single-cell RNA sequencing (scRNA-seq)  has provided comprehensive information about transcriptomes, altering our ability to identify cell subpopulations. However, the segregation of cells while dissociating the tissue destroys cellular spatial information in the original tissue context, which sometimes could be extremely crucial to understanding intricate cellular interaction networks. Moreover, since scRNA-seq was developed in 2009, many limitations have been emerging. For instance, the relatively low efficiency and coverage of RNA transcript capturing may lead to the loss of gene expression information for downstream analysis . Furthermore, certain types of cells may exhibit significant cell variations due to factors such as cell size and cell cycle stage, causing less reliable results. Another challenge of scRNA-seq is the batch effect which also needs to be considered and corrected before subsequent analyses . Additionally, the dissociation protocol of tissue sections may have repercussions on transcriptome and induce transcriptome-wide changes including ectopic expression of genes, causing a contaminating signal and subsequently leading to the misidentification of cell subpopulations . These obstacles are gradually improved with advances in spatial transcriptomics where each cell is assigned a specific and unique spatial label containing spatial coordinates information, allowing for relatively precisely positioning each identified cell subpopulation to the original tissue sections . Employing spatial transcriptomics techniques enables transcriptomic data to be acquired from intact tissue sections and in turn obtains spatial distribution information and elucidates cellular interaction patterns .
Although current cutting-edge spatial transcriptomics techniques are confronted with some drawbacks such as relatively low resolution and comparatively insufficient sequencing depth , they are extensively utilized in a wide range of biomedical research because of the accelerating capacity to investigate the spatial architecture of normal tissue and tumor. These approaches and platforms have been applied to the adult mouse brain , mouse liver , human dorsal root ganglia  and dorsolateral prefrontal cortex , human heart , embryonic liver , intestine  and mammalian testis  to reveal tissue architecture and delineate embryonic developmental blueprint and also been employed to lucubrate disease pathogenesis and microenvironment [14,15,16,17]. An important part of the disease research is into tumor biology which encompasses pancreatic ductal adenocarcinoma , human squamous cell carcinoma , breast cancer  and cutaneous malignant melanoma , etc. These applications provide adequate novel biological insights and clinical relevance to resolving the intrinsic mechanism of tissue dynamics and disease and to remedying or optimizing present medical treatment protocols. Bioinformatics analysis strategies aim at mutual and disparate purposes concerning clustering analysis, data integration, deconvolution, spatially-variable genes identification, etc. For example, early-developed and now commonly-used Seurat  can be applied to clustering and gene imputation, and the recently published Tangram  tackles deconvolution and also gene imputation.
Spatial transcriptomics technologies have been continuously making significant progress. Multiple technologies have emerged in recent years, and their applications and advantages and disadvantages are comprehensively reviewed. In this article, we summarize the landscapes of available spatial transcriptomics technologies, present the employment of spatial techniques in extensive fields of biomedical research and focus on the status quo of computational strategies of data analysis.
Development of spatial transcriptomics technologies
Since the initial spatial transcriptomics workflow was established in 2016 , this field has been proceeding apace with the unceasing evolution in resolution as well as throughput. Notably, spatially resolved transcriptomics was heralded as “Method of the Year 2020” by Nature Methods in 2021 . Feasible methods for obtaining a fine-grained assessment of spatial transcriptome can be generally classified into four primary categories including microdissection, in situ hybridization, in situ sequencing, and spatial barcoding, each bearing its superiority and constraints. Overviews of these categories are summarized and a concise timeline depicting the remarkable course of spatial transcriptomics techniques is presented (Fig. 1) and detailed comparisons among existing methods are shown (Table 1). Some of the most commonly used spatial transcriptomics platforms are also listed in Table 2.
Technologies based on microdissection
Laser capture microdissection (LCM)  is a microdissection technique that employs a focused infrared laser pulse to isolate a specific tissue region of interest (ranging from 60 to 700 μm in diameter) from the original tissue section, enabling precise procurement of a specimen from the specified anatomical region while diminishing potential contamination. Moreover, these technologies are appropriate for partly-degraded tissue section analysis  and can interrogate the transcriptomes at a cellular resolution. One application of LCM technology is the genetic analysis of small premalignant lesions that have been isolated from histologically normal tissue or tumor edges, and this approach underlies several other technologies including tomo-seq , Geo-seq , etc.
Junker and colleagues  devised RNA tomography (tomo-seq), a technique that involves cryosectioning, reverse transcription, and amplification. Notably, this approach eliminates the need for carrier RNA and provides high sensitivity and spatial resolution. The robustness of the tomo-seq protocol was validated by the authors by applying it to zebrafish embryos, followed by a three-dimensional reconstruction of a genome-wide atlas at three developmental stages of the zebrafish embryo. The 3D profiling of tomo-seq was accomplished by cryosectioning three main body axes of the zebrafish and the data sets measured along these axes were reconstructed computationally by mapping gene expression information onto the image. Analysis of the 3D transcriptomic pattern of whole embryos and organs can be accomplished by tomo-seq but a main drawback of this method is that multiple samples are needed to generate sections of three axes so the application on human organs can be limited. Chen and colleagues  proposed another technology based on microdissection termed geographical position sequencing (Geo-seq) which integrates LCM and scRNA-seq technologies, enabling simultaneous investigation of cell heterogeneity and spatial variation. Geo-seq implements gene profiling at a ten-cell resolution, significantly facilitating the analysis of the spatiotemporally-regulated gene expression compared to individually utilizing the LCM method. In addition, Geo-seq can also promote the understanding of rare cells and the interaction between cells and surrounding niches. However, some impediments still remain, including the amplification merely of mRNA with a poly-A tail while preparing the library, which can be a hindrance for the subsequent Smart2-seq .
In summary, microdissection-based methods provide a competent approach to obtaining regions of interest from tissue samples with high sensitivity. These techniques enable focused research into the microanatomical structures and gene expression information of specific regions. However, Geo-seq, which integrates LCM and scRNA-seq (Smart2-seq), offers only a ten-cell resolution due to the limitations of microdissection-based techniques. During the laser-capturing and tissue segregation procedures of LCM, the quality of RNA molecules and the intactness of obtained cells may not be fully maintained. Additionally, microdissection is time-consuming and labor-intensive, limiting the throughput and the capacity to handle large tissue samples. Despite these shortcomings, microdissection-based technologies can still provide robust methods for gene expression profiling.
Technologies based on in situ hybridization
In situ hybridization is a strategy that enables the visualization of RNA molecules within their original context via probes complementary to the objective transcripts rather than extracting them from tissue sections. An early iteration of in situ hybridization technique termed single-molecule fluorescent in situ hybridization (smFISH)  is competent in detecting several RNA transcripts simultaneously and has been advancing in gene measuring throughput and efficiency through multiplexed smFISH [30, 31]. This method exhibits high sensitivity and offers a subcellular resolution and is commonly utilized as a powerful tool for biological validation, such as corroborating the findings of bioinformatic analyses for newly identified genes. This technology requires fluorescent labeled RNA probes to hybridize with target molecules so the main drawback of smFISH is the limitation on the number of color channels due to the fluorescent overlapping of different channels, which means that smFISH can detect only a small number of genes concurrently. Another in situ hybridization technology called ouroboros smFISH (osmFISH)  is a non-barcoded and unamplified method based on cyclic smFISH, which can identify weakly-expressed genes  due to the circumvention of optical crowding. OsmFISH can be applied to large tissue samples, particularly for the examination of low-expression RNA transcripts. However, low throughput remains a technical limitation of this technique. Sequential FISH (seqFISH) is a barcoding protocol that leverages the high efficiency of FISH and the fact that distinguishing RNA transcripts does not require base-pair resolution . In this approach, mRNAs are assigned temporal barcodes through multiple rounds of hybridization. During each round of hybridization, each transcript is targeted with several probes labeled with one color, and subsequently the probes are removed before the next round of hybridization where the same probes are labeled with fluorophores of a different color. Thus, seqFISH can generate a large number of transcripts while reducing spectral overlap that occurs in smFISH. However, seqFISH can be time-consuming and errors may accumulate over multiple rounds of hybridization, potentially leading to inaccurate information. Despite these limitations, seqFISH can be used to generate transcriptomic images of complex tissues, including brain samples .
To overcome the drawbacks of accumulating errors, Chen and colleagues  devised multiplexed error-robust FISH (MERFISH), a highly multiplexed smFISH protocol incorporating combinatorial labeling, successive rounds of sequential hybridization imaging, and error-robust encoding. MERFISH workflow is capable of measuring genes and combating accumulating detection errors by the error-robust encoding strategy designating each RNA transcript with a binary word. A 140-gene measurement was simultaneously performed with the encoding strategy that can detect and correct errors, whereas a 1001-gene measurement was performed with an alternative encoding strategy which can detect errors, albeit with no correction . Notably, efforts have been made to evolve the MERFISH approach, enabling the simultaneous detection of RNA molecules to achieve up to 10,000 . Moreover, MERFISH can be implemented to accomplish a high-throughput analysis of intercellular gene expression variation and elucidate the spatial distributions of multiple RNA transcripts concurrently. In contrast to seqFISH, the MERFISH protocol removes fluorophores but not the probes, making it more time-efficient than seqFISH . The MERFISH approach has been commercialized as Vizgen MERSCOPE (Table 2) and can be applied to multiple tissue samples including fresh frozen and formalin-fixed paraffin-embedded (FFPE) tissue sections.
Overall, in-situ-hybridization-based techniques allow for the visualization of RNA molecules within their original tissue context by hybridizing probes with complementary targets. This enables the detection of target genes for biological validation of bioinformatic analysis results and the study of gene expression patterns. However, the nature of FISH methods imposes an intrinsic limitation on throughput. Additionally, specific probes must be synthesized before the hybridization process, necessitating the use of ready-made kits to overcome this challenge .
Technologies based on in situ sequencing
In situ sequencing (ISS) method developed by Ke and colleagues  enables targeted analysis of RNA molecules in cells within a histomorphologically-retained context. This protocol entails single-strand DNA padlock probes with complementary sequences that bind to the cDNA generated by reverse transcription of mRNA molecules. Two targeted approaches, gap-targeted sequencing and barcode-targeted sequencing, were developed in the ISS procedure. In gap-targeted sequencing, the padlock probe has a gap between the probe ends which precisely binds to the targeted base pairs in the cDNA, and DNA polymerization and ligation subsequently fill the gap to form a circular DNA molecule. In barcode-targeted sequencing, the padlock probe contains a barcode sequence and only one breakpoint, so the formation of circular DNA undergoes only the ligation of the breakpoint. Rolling-circle amplification of the circularized DNA generates a rolling-circle product which then undergoes sequencing by ligation. The accuracy of the ISS protocol has been validated through its implementation in human breast cancer to manifest point mutations and decompose multiplexed gene expression profiling, using gap-targeted sequencing and barcode-targeted sequencing, respectively . However, the ISS method requires prior knowledge of examined tissue to design padlock probes.
To examine transcripts without prior knowledge of tissue, Lee and colleagues  devised fluorescent in situ RNA sequencing (FISSEQ), a non-targeted approach measuring 8102 RNA species unbiasedly (transcriptome-wide). FISSEQ predominantly detects genes depicting cell type and function but low sequencing depth and incapability of ascertaining targeted RNA remain to be the drawbacks. Based on FISSEQ, another in situ sequencing strategy named expansion sequencing (ExSeq) was launched, enabling highly-multiplexed RNA visualization in cells and tissues of multiple-organ species with high spatial precision . ExSeq encompasses targeted and untargeted versions, both of which can resolve biological problems ranging from nano-scale to system-scale. The targeted version addresses the issue of cellular crowding by attaching RNA molecules to an expandable hydrogel and expanding the hydrogel before ligating and sequencing, and the untargeted version optimizes the efficiency . Untargeted ExSeq allows the detection of RNA molecules in the whole transcriptome including rare transcripts, whereas targeted ExSeq enables a smaller defined gene set to be detected and can be utilized to project cells onto tissue context and also visualize gene regulation. Wang and colleagues  developed spatially-resolved transcript amplicon readout mapping (STARmap) incorporating hydrogel-tissue chemistry and in situ sequencing, which can be employed to sequence RNA in 3D intact tissue with high efficiency and accuracy. Additionally notably, a modified STARmap scheme can be adopted for 3D analysis of thick tissue blocks, and sequencing with error-reduction by dynamic annealing ligation (SEDAL) was specifically devised for STARmap to eradicate misdecoding resulting from sequencing errors.
In contrast to traditional sequencing methods that separate cells from their spatial context, in-situ-sequencing-based methods enable spatial-level gene expression analysis and avoid the bias introduced by transcript extraction. However, these techniques still face challenges. For example, prior knowledge of the tissue may be required to design specific padlock probes, and read length may be limited. Additionally, in situ sequencing may not be feasible for unconventional or rare cell types and genes. Potential applications of these methods include studying gene expression regulation within tissues or cells and localizing gene variants.
Technologies based on spatial barcoding
Ståhl and colleagues  proposed Spatial Transcriptomics (ST), which is practicable for quantitatively visualizing and determining the transcriptome whilst retaining spatial information. Tissue sections of adult mouse olfactory bulbs are placed on the glass slides immobilized with reverse transcription primers with poly-T to bind to the poly-A tail of mRNA derived from the tissue sections. The primers also embody spatial barcodes and unique molecular identifiers (UMIs) representing the coordinates of each array. During the tissue permeabilization process, mRNA molecules in tissue cells diffuse into 100-μm microwells on slides and hybridize with primers. Reverse transcription reagents are then added to the tissue to synthesize cDNA, using Cy3-labeled nucleotides for visualization of the generated cDNA. The tissue is subsequently removed by enzymes, leaving cDNA hybridized with nucleotides on the glass slides . Although this technology provides spatial information, the resolution is limited to 100 μm, containing multiple cells. In 2019, 10 × Genomics further developed this method and commercialized it as “10 × Genomics Visium”, upgrading the resolution to 55 μm and refining the protocol to be compatible with both fresh frozen tissue sections and formalin-fixed paraffin-embedded (FFPE) tissue sections. This method has been widely used to study various tissue and disease. Maynard and colleagues  initially exploited the Visium platform to interpret gene expression information spatially in the human DLPFC on a transcriptomic scale.
Improvement of the resolution of spatial barcoding strategies has been continuously pursued. In 2019, Rodriques and colleagues  developed Slide-seq which provided an approach for spatially analyzing gene expression information at high resolutions (10 μm) analogous to the size of a single cell using beads deposited on the slide, with scalability to the large volume of tissue. Since these beads are randomly placed on the slide surface, their position information must be decoded through sequencing to match transcripts with their location, which may limit the capture efficiency. In 2021, Stickels and colleagues  described the improved version of Slide-seq, termed Slide-seqV2, which advanced approximately an order of magnitude in RNA capturing efficiency and sensitivity than the original Slide-seq. Not long after the publication of Slide-seq, a high-resolution spatial technology named high-definition spatial transcriptomics (HDST) utilizing barcoded bead arrays to capture RNA molecules from tissue sections in a histological context achieved a 2-μm resolution which is much higher than Spatial Transcriptomics . It is also prominent that Seq-Scope technology yields a submicrometer resolution of 0.5 ~ 0.8 μm .
Slide-seq, HDST and Seq-Scope introduced above can provide much higher and even subcellular resolutions, generating more refined spatial distribution information. The approaches to improving the resolutions of Slide-seq and HDST are similar, involving bead arrays with 10-μm- and 2-μm-diameter beads, respectively [43, 45]. It should be noticed that Slide-seq and HDST involve beads similar to or smaller than the size of a single cell but they may cover multiple cells so the single-cell resolution may not be always achieved. Seq-Scope achieves subcellular resolution through the dense distribution of clustered barcodes. To be specific, many oligonucleotides containing high-definition map coordinate identifiers (HDMI) act as seed molecules, and an HDMI-array is generated by amplifying these seed molecules to form many clusters, each of which is derived from one seed molecule. This process can almost eliminate the areas with no detected RNA molecules . However, pursuing such high resolution may introduce challenges such as data sparsity and difficulty inferring cell borders . Noise is also a challenge due to limited coverage in each sequencing unit and the complex procedures required to maintain spatial positions during sequencing. The higher the resolution is, the more severe the noise is likely to be . To improve the resolution while preserving comprehensive and necessary information, future breakthroughs may involve smaller but more sensitive detection units and the integration of spatial transcriptomics with high-throughput scRNA-seq data.
Overall, spatial-barcoding-based techniques allow for the simultaneous acquisition of gene expression and spatial location information. However, selecting the appropriate resolution requires careful consideration. Low resolution may obscure the intrinsic tissue structure and require further decomposition analysis to gain comprehensive insights, while high resolution may introduce those aforementioned challenges. Additionally, capture efficiency may be relatively low. Despite these limitations, spatial-barcoding-based techniques are widely used to study tissue architecture, tumor heterogeneity, the tumor microenvironment, etc.
Gaining biological insights from spatial transcriptomics
Spatial transcriptomics technologies are potent tools for studying the intricate structure, the dynamics of tissue and organ systems and inherent mechanisms within their original context. These technologies can provide valuable biological insights by revealing tissue architecture, developmental patterns and diseases, among which tumor biology may be one of the most extensive applications of spatial transcriptomics. Primary application scenarios of implementing spatial transcriptomics techniques are presented (Fig. 2) and several representative studies utilizing spatial transcriptomics are enumerated (Table 3).
Illustrating tissue architecture and developmental atlas
Decoding intercellular interaction and identifying cell subpopulations are of fundamental significance in delineating tissue architecture and defining structural components through the establishment of a transcriptome atlas of a specific tissue or organ, thus facilitating the perception of tissue dynamics. Hildebrandt and colleagues  managed to delineate the transcriptional landscape of sectioned mouse liver by employing spatial transcriptomics, corroborating the concept that liver lobular zonation characterized tissue heterogeneity by profiling of pericentral and periportal expression of representative marker genes. Ortiz and colleagues  accomplished a molecular atlas by applying spatial transcriptomics to a whole mouse brain to spatially manifest the brain tissue organization and composition. They also used a scRNA-seq dataset containing both neuronal and nonneuronal cells to map their spatial positions using a trained neural network model. This study demonstrates the potential of spatial transcriptomics to analyze complex samples such as brains, in addition to other tissues or organs. In addition, a study on the human dorsolateral prefrontal cortex (DLPFC) also resorts to spatial transcriptomics, which is notably the first research adopting the 10 × Genomics Visium platform, the commercialized version of spatial transcriptomics . This study demonstrates the transcriptome-wide gene expression topography of human DLPFC across cortical laminae and subsequently a series of bioinformatics analyses are conducted to refine previous lamina-enriched genes and identify novel lamina-enriched genes. Moreover, the study delves into schizophrenia and autism spectrum disorder by incorporating previously-procured publicly-available neuropsychiatric disorder gene datasets to distinguish the particular lamina where genes associated with the diseases enrich, underlining the clinical significance of the study. Another study utilizing 10 × Genomics Visium probes into human nociceptors to present molecular features by applying the technology to human dorsal root ganglia . Given that nociceptors are principal targets for acute and chronic pain treatment, the study might also provide insights into advancing medical treatment protocols and identifying novel drug targets.
Furthermore, spatial transcriptomics technologies are generally utilized in developmental biology to reveal spatiotemporal gene expression patterns and uncover tissue morphogenesis throughout the entire development course or multiple pivotal stages. Asp and colleagues  profiled a cell atlas of human cardiogenesis course where three developmental stages of the human embryonic heart were comprehensively delineated. They combined spatial transcriptomics with scRNA-seq to perform single-cell analysis and identify multiple cell types, and exploited in situ sequencing to position cells within their original clusters. The integration of spatial transcriptomics, scRNA-seq and in situ sequencing provides comprehensive insights into spatiotemporal patterns, marker genes, cellular interaction networks and developmental trajectories. Chen and colleagues  generated a spatial atlas for the transcriptome of mammalian spermatogenesis by adopting Slide-seq to mouse and human testis specimens and further characterized the microenvironment surrounding and mediating spermatogonial course by combining in situ sequencing.
Beyond the above insights about tissue architecture and development, spatial transcriptomics techniques have a robust capacity for clarifying disease microenvironments and pathogenesis. Boyd and colleagues  combined scRNA-seq with spatial transcriptomics to interrogate tissue inflammatory impairment in acute respiratory distress syndrome induced by severe respiratory influenza A virus infections. Their findings provided compelling evidence of the essential role played by lung fibroblasts in regulating immune reactions at the site of infections. This study demonstrates the utility of spatial transcriptomics in studying inflammatory diseases and the immune microenvironment and has stimulated research into immunopathy of other infectious diseases, including COVID-19, which continues to be a global health concern. Maniatis and colleagues  employed spatial transcriptomics on spinal cords from mice and amyotrophic lateral sclerosis patients to gain gene expression information to elucidate spatiotemporal dynamics mediating the degeneration of motor neurons. This research identifies the locations and distributions of specific genes associated with the disease and elucidates the underlying mechanisms regulating this neurodegenerative disorder.
A substantial part of disease research is the study of tumor biology which could be the most extensive application of spatial transcriptomics. Significant challenges in devising tumor treatment procedures are induced by tumor heterogeneity. Moncada and colleagues  utilized both scRNA-seq and spatial transcriptomics to investigate pancreatic ductal adenocarcinomas and distinguished cell populations and subsequently generated an unbiased map of the transcriptomes across the tumor, revealing its intrinsic architecture and heterogeneity. Another study that combined scRNA-seq and spatial transcriptomics to delineate the constitution and spatial architecture of cells within cutaneous squamous cell carcinoma revealed the cancer cell subpopulations and their communication . The tumor microenvironment has become another hotspot of tumor-related research due to its complexity and diversity. Deciphering the tumor microenvironment is crucial for perceiving the intricate interactions between the tumor and microenvironment and may also aid in tumor immunotherapy. One study integrating spatial transcriptomics and scRNA-seq revealed the tumor microenvironment related to the immunotherapeutic efficacy of hepatocellular carcinoma, demonstrating a potential treatment target . Another study analyzed the interactive relationship between the host and the microbiota in oral squamous cell carcinoma and colorectal cancer at a spatial level utilizing spatial transcriptomics and GeoMx digital spatial profiling . It indicated that the tumor-associated microbiota, as an essential part of the tumor microenvironment, could impact tumor heterogeneity and induce the migration of cancer cells. Wong-Rolle and colleagues  conducted research related to intratumoral bacteria, where they discovered the enrichment of intratumoral bacteria in lung cancer and their association with several oncogenic pathways. The employment of spatial transcriptomics in tumor biology can reveal tumor heterogeneity and microenvironment to a large extent, thus providing ample instructions on addressing current obstructions confronting the treatment protocols.
Data analysis of spatial transcriptomics
To comprehensively interrogate the tissue sections, bioinformatic analyses have to be performed to unravel the intertwined and multiplexed bioinformation and minimize the impact of current technological limitations and subsequently derive biological significance more accurately from raw spatial transcriptomics data. These bioinformatics analyses range from spatially-variable genes identification and clustering analysis to gene imputation, etc., which can be handily effectuated through a substantial number of computational strategies devised in recent years. Herein, circumstantial comparisons of algorithms and usages among the existing R or Python packages are presented (Table 4).
Distinguishing cell types and subpopulations is a fundamental task in the bioinformatic analysis of spatial transcriptomics data. This can be resolved with the help of clustering analysis where spatially-variable genes can be discovered and data dimensions can be reduced through approaches such as principal component analysis (PCA), t-distributed stochastic neighbour embedding (t-SNE) and uniform manifold approximation and projection (UMAP). These methods calculate similarity among barcode spots and define clusters within a tissue. A robust clustering procedure is provided by a widely-distributed R package Seurat , on which another R package capable of clustering analysis STUtility builds its framework . Seurat is prevalent in scRNA-seq and spatial transcriptomics data analysis and is also competent in other bioinformatics analyses such as gene imputation. Zhao and colleagues  proposed BayesSpace based on a Bayesian model with a Markov random field, which outperformed previous clustering algorithms and improved spatial transcriptomics resolution to subspot levels. BayesSpace was validated by analyzing tissue samples, including brain and melanoma, overcoming challenges of low resolution and technical noise. SpaGCN is a python package based on a graph convolutional network that incorporates gene expression, spatial coordinates, and tissue histology visualization . Clustering analysis is accomplished by aggregating gene expression from neighboring spots using a graph convolutional layer. SpaGCN has been tested on various species and utilized to analyze data generated from Spatial Transcriptomics and MERFISH. However, this strategy has the limitation of potential disagreement between actual tissue structure and detected spatial regions because the detection of spatial regions is primarily driven by gene expression information.
Spatially-variable genes identification
Within a certain tissue, some genes exhibit conspicuous spatially-variable expression whereas some other genes such as housekeeping genes are expressed equally among the cells. The specific pattern in which the expressions of genes spatially vary can convey indispensable bioinformatic insights into identifying cell types and subpopulations and corresponding spatial information and underlying spatial functions. Some program packages perform outstandingly in identifying spatially-variable genes. Svensson and colleagues  described a strategy named SpatialDE, based on Gaussian process regression, which utilized two random effect models including a spatial variance model and a noise model to decompose variable expression of each gene into spatial and non-spatial components, respectively. Another package that identifies genes with statistical significance in spatial expression is termed trendsceek, building on marked point processes . The trendsceek strategy can be performed on spatially resolved transcriptomics data sets and also scRNA-seq data projected onto a low dimension. Spatial pattern recognition via kernels (SPARK) technology, based on a generalized linear spatial model with a penalized quasi-likelihood algorithm, can overcome the high type I errors and low statistical power of previous strategies such as SpatialDE and trendsceek and is furthermore capable of analyzing large-scale spatial transcriptomics datasets . However, SPARK may perform better for certain datasets and genes, causing intrinsic bias.
Spatial decomposition and gene imputation
A common issue in spatial transcriptomics technology is that a single barcode-capturing spot may be overlaid by multiple cells. Thus, the detected expression is an aggregation of a heterogeneous set of cells within the spot, which may impact the efficiency and accuracy of identifying cell subpopulations and delineating tissue atlas. For example, 10 × Genomics Visium offers a resolution of 55 μm meaning the diameter of each capturing spot is 55 μm which is several-fold larger than a typical tissue cell. The spatial decomposition process through various deconvolution algorithms can address this discrepancy, which is to disentangle the mixture of mRNAs and subsequently predict the proportions of each cell type in one capturing spot. A spatial decomposition method devised by Ma and colleagues  is termed conditional autoregressive-based deconvolution (CARD) building on a non-negative matrix factorization model, which outperforms SPOTlight , RCTD , SpatialDWLS , etc. in deconvolution accuracy, corroborated by correlation analysis with scRNA-seq data. One potential improvement to this strategy is to incorporate tissue images, allowing for easier comparison between histological features and analysis results.
Gene imputation refers to the task of inferring lost gene expression information or “dropouts” caused by factors such as low protocol sensitivity, mitigating errors during gene measurement and facilitating deconvolution. Biancalani and colleagues  introduced a deep learning framework Tangram performing gene imputation. Gene imputation generated by Tangram yields an estimation of “dropouts” and prediction of spatial expression patterns more accurately conforming to MERFISH technology which is also competent in combating detection errors , thus promoting deconvolution of cells hampered by “dropouts”. The integrative and widespread R package Seurat can also impute gene expression utilizing co-expression patterns . Abdelaal and colleagues  proposed Spatial Gene Enhancement (SpaGE) incorporating scRNA-seq and spatial data to predict gene expression which spatial transcriptomics techniques fail to detect, depending on a domain adaptation model. SpaGE is flexible and scalable when applied to large datasets and outperforms previous tools.
The aforementioned strategies, including spatial decomposition and gene imputation, have demonstrated considerable efficacy in enhancing the resolution of spatial transcriptomics data and compensating for lost gene expression information. Nevertheless, certain limitations persist. These approaches are based on computational models for predicting cell locations and gene information and therefore, their predictions may be subject to error, potentially resulting in imprecise and spurious results. Further investigation and refinement are necessary to more effectively leverage these technologies and derive more reliable biological insights.
Cellular interaction operated within the microenvironment where cells are adjacent to each other can convey significant perceptions into tissue dynamics and the way the communication networks change when experiencing conditions such as disease. A Graph Convolutional Neural networks for Genes (GCNG) method was introduced to infer extracellular interactions from gene expression by depicting a cellular relationship graph transformed from spatial transcriptomics data and subsequently encoding gene expressions, and the graph is then convolved with expression information . Cang and colleagues  launched spatially optimal transporting the single cells (SpaOTsc) to obtain intercellular communication, based on a structured optimal transport model. However, SpaOTsc does not account for time delays during intercellular communication. Owing to the three-dimensionality of tissue blocks, utilizing exclusively either scRNA-seq or spatial transcriptomics cannot output sufficient information to decipher cellular communication networks, therefore the integration of both datasets becomes a fundamental consideration when conducting bioinformatic analysis.
Spatial copy number variations identification
Copy number variation (CNV) refers to the increase or decrease in the copy number due to gene segment rearrangements. Typically, CNVs involve segments longer than 1000 base pairs and are mainly manifested as submicroscopic deletions or duplications. CNVs are a common form of genetic variation in the human genome, with 5% ~ 10% of the genome affected by CNVs, which is much higher than other forms of genetic variation. Ascertaining the transition from benign to malignant tissue forms the foundation for improving early cancer diagnosis, as genomic instability in histologically benign tissue can signal an early event in cancer evolution. Furthermore, the spatial distribution and activity of CNVs can impact phenotype, making mapping their spatial distribution valuable for comprehending, diagnosing, and treating diseases. Previously, gene expression was utilized to infer CNVs in individual cells, successfully identifying regions of chromosomal gain and loss . Erickson and colleagues  expanded this approach to a spatial modality with the development of SpatialInferCNV, an R package that identifies CNVs in each spatially barcoded region. Additionally, another package named SPATA also integrated a module for CNV detection .
Region annotation and spatial trajectories
Gene expression within a tissue is influenced by the spatial position of cells in the tissue microenvironment. Spatial transcriptomic data can provide valuable insights into tissue regions, as they contain information on spatial position matrices, HE region staining of sections, and relative distances between individual cells, which can be used to delineate spatial regions. MULTILAYER is an algorithm that utilizes agglomerative clustering and community detection methods for graphical partitioning, enabling digital imaging of spatial transcriptomic analysis . This allows for contextual gexel (namely, the locally defined transcriptomes) classification strategies, which can be used to develop self-supervised molecular diagnosis solutions.
Spatial trajectory analysis is an analytical method frequently employed in spatial transcriptomics to uncover dynamic cellular evolution and differentiation processes. This approach infers evolutionary trajectories and differentiation relationships between cells by analyzing their spatial positions and gene expression levels within tissue sections. The stLearn package can visualize spatial trajectories in tissue slices and infer biological processes from transcriptional state gradients across tissues . Similarly, SPATA concentrates on temporal alterations in gene expression to deduce transcriptional patterns dynamically governed by the spatial organization .
Both spatial transcriptomics and scRNA-seq are effective methods for obtaining biological insights into tissues and diseases. However, each method has its limitations. By integrating spatial transcriptomics and scRNA-seq data, these methods can complement each other to provide comprehensive biological information. For instance, RCTD generates spatial decomposition by assigning cell types to spatial transcriptomics spots , whereas Tangram performs gene imputation by aligning scRNA-seq data with spatial transcriptomics data to learn spatial transcriptome-scale paradigm . Additionally, CellTrek is a computational strategy that integrates scRNA-seq and spatial transcriptomics data sets to perform spatial decomposition by reconstructing a cellular map on tissue sections . This strategy is distinct from other spatial decomposition methods in that CellTrek directly maps single cells to corresponding spatial positions in the spatial context. Other than these R or Python packages, many studies have incorporated spatial transcriptomics and scRNA-seq. Liu and colleagues  discovered a tumor immune barrier structure and a series of cancer-associated fibroblasts related to the efficacy of immune treatments through an integrative analysis of spatial transcriptomics and scRNA-seq. The scope of ‘data integration’ encompasses not only the alignment of these two methods but also the incorporation of spatial transcriptomics with other omics data. However, few individual computational tools are designed specifically for combining spatial transcriptomics and other omics. Therefore, linking multiple packages for analysis is necessary. For instance, a remarkable study integrated spatial transcriptomics, scRNA-seq, proteomics and whole-exome sequencing to resolve pancreatic cancer microenvironment, utilizing various packages including Seurat, RCTD, CellPhoneDB (for detecting ligand-receptor interactions), Monocle3 (for inferring cell transitions), inferCNV (for detecting CNVs in scRNA-seq data), germlinewrapper and somaticwrapper (for calling germline variants and somatic variants, respectively), among others . Thus, we can see the significant potential in the integrative analysis of spatial transcriptomics, scRNA-seq and other omics.
A brief pipeline of spatial transcriptomics data analysis
Methods for analyzing spatial transcriptomics data are generally similar and can be divided into data preprocessing and downstream analysis. Data preprocessing typically involves quality control and normalization to improve data quality for downstream analysis and obtain more reliable biological information. For spatial-barcoding-based methods, quality control aims to remove low-quality spots and genes from spatial transcriptomics data. Quality control parameters can be adjusted based on tissue type, research requirements, and other factors. These parameters may include removing spots with fewer than a certain number of transcripts, removing genes expressed in fewer than a certain number of spots, and removing spots with a high proportion of mitochondrial genes. Normalization accounts for the difference in sequencing depth among different spots. Since differences among spots in spatial transcriptomics data can be relatively large, effective normalization is essential.
After preprocessing, downstream analysis can be performed. The data should first undergo dimensionality reduction and clustering analysis to distinguish spots with different features. Biological information can then be interpreted through these clusters in subsequent analysis. Algorithms such as PCA, t-SNE, and UMAP can be used for this purpose and are available in many data analysis packages. Next, gene expression patterns in the data can be analyzed, including differential expression analysis and spatially variable gene analysis, which can be performed using packages such as Seurat and SpatialDE, respectively. Additionally, cell information from tissue slices can be annotated onto spatial transcriptomics data. Since the sequencing unit (e.g., spots in 10 × Genomics Visium and beads in Slide-seq) of some spatial transcriptomics technologies may contain more than one cell, spatial decomposition can infer the proportion of various cells in each sequencing unit based on the data to obtain cell locations in the spatial context. This step can be achieved using packages with deconvolution algorithms such as RCTD and cell2location. Gene imputation can also predict the positions of low-expressed or missing genes in space due to possible dropout using packages like Tangram. Furthermore, personalized analysis can be conducted based on research objectives. For instance, packages such as Giotto can be used to analyze the communication between cells or spatial regions, including receptor-ligand interactions. SpatialInferCNV can perform copy number variation analysis at the spatial level, while stLearn and SPATA can be used for spatial trajectory analysis and MULTILAYER for spatial region identification. These analytical methods and packages provide excellent visualization during data analysis, facilitating step-by-step comprehension of current analytical outcomes to guide subsequent analysis. Moreover, it is essential to integrate spatial transcriptomics data with scRNA-seq data and other omics data to obtain a more comprehensive understanding of biological information.
Conclusion and future perspectives
Explosive advances in spatial transcriptomics technologies have been made in recent years to expand our understanding of miscellaneous tissues and organs. However, current spatial transcriptomics methods are confronted with some challenges of low resolution, sensitivity, throughput, etc., hindering our precise perception of normal and abnormal tissues, which calls for further innovations in technologies to overcome these deficiencies. Given that each technology bears its biological strengths, we envision the integration across these technologies which complement each other in the drawbacks before a novel and robust technology is launched. With future technology revolutions, intercellular signaling could be resolved at higher and even single-cell resolution. In addition, larger-scale tissue specimens may be investigated to allow for depicting organ-level tissue topography, enabling a more holistic and consecutive interpretation of tissue structures, which latently poses challenges for accelerating bioinformatic analysis with higher efficiency and accuracy and more powerful information processing capacity. Beyond the prospective advancement in refining and optimizing current protocols of spatial transcriptomics, we also envisage the integration with multi-omics including epigenomics, proteomics, and metabolomics to shed light on the intrinsic convoluted mechanisms of cellular interactions and disease and better probe into tumor progression and growth course. In addition to advances in spatial transcriptomics technologies, innovations in data analysis strategies are also anticipated. As deep learning technology continues to progress, its application in spatial transcriptomics data analysis is expected to become more widespread. In the future, more deep-learning-based methods may be developed to process and analyze spatial transcriptomics data to improve data resolution and interpretation reliability. Furthermore, as data scale and complexity increase, visualization and interactive analysis will become important tools for spatial transcriptomics data analysis. Future spatial transcriptomics data analysis methods will need to integrate visualization and interactive analysis technologies to better understand and interpret data.
Since some spatial transcriptomics techniques, especially some widespread spatial-barcoding-based techniques, are not capable of offering single-cell resolution at the spatial level and scRNA-seq cannot reflect the spatial distribution of each cell, we envision a more organic and efficient alignment of single-cell datasets and corresponding spatial information. The alignment can be achieved by mapping single cells to spatial data, where each cell is matched with a spatial location in an ideal condition. Nevertheless, current methods for integration cannot generate precise matching due to technological limitations, which calls for further breakthroughs in the effectiveness and efficiency of data integration algorithms. By integrating both datasets, we can decipher potential intercellular communication pathways, including ligand-receptor interactions and juxtacrine and paracrine signaling. This may provide insights into previously unclear physiological and disease mechanisms and help discern more refined classifications of certain diseases, facilitating precise and individualized medical treatment. Additionally, publicly-available datasets can be interrogated retrospectively with the integration of spatial transcriptomics and scRNA-seq data to obtain novel biological cues which may be concealed in the raw data before.
Moreover, we anticipate the translational medicine research into the clinical significance of spatial transcriptomics, particularly with the compatibility of the 10 × Genomics Visium platform with FFPE tissue blocks allowing retrospective analysis into previously opaque tissue specimens to glean more sufficient information on clinical diagnostics and prognostics as well as therapeutic methods and targets. For example, research into human DLPFC distinguished the layer-enriched genes that may be associated with schizophrenia and autism spectrum disorder, implicating the potential of neuropsychiatric disorders progression in those bearing the risk gene expression . In tumor biology, spatial transcriptomics incorporated with other omics can identify cancer gene signatures and subsequently reveal novel targets for cancer treatment and assist us to abate or suppress the degree of tumor cell proliferation, infiltration, and invasion. Nevertheless, it is noteworthy that before translating omics data into clinical relevance, the robustness of the technologies and the quality of specimens and specimens processing must be considered.
Availability of data and materials
Single-molecule RNA fluorescence in situ hybridization
Laser capture microdissection
Single-cell RNA sequencing
In situ sequencing
Transcriptome in vivo analysis
Fluorescent in situ RNA sequencing
Sequential fluorescence in situ hybridization
Multiplexed error-robust fluorescence in situ hybridization
Single-molecule hybridization chain reaction
Geographical position sequencing
Barcode in situ targeted sequencing
Spatially-resolved transcript amplicon readout mapping
Ouroboros single-molecule RNA fluorescence in situ hybridization
Digital spatial profiling
High-definition spatial transcriptomics
Deterministic barcoding in tissue
Spatial enhanced resolution omics-sequencing
Expansion spatial transcriptomics
Unique molecular identifiers
Dorsolateral prefrontal cortex
Corona virus disease
Dorsal root ganglia
Amyotrophic lateral sclerosis
Influenza A virus
Benign prostatic hyperplasia
Pancreatic ductal adenocarcinoma
Cutaneous squamous cell carcinoma
Conditional autoregressive-based deconvolution
Deconvoluting spatial transcriptomics data through graph-based convolutional networks
Graph convolutional neural networks for genes
Robust cell type decomposition
Spatial transcriptomic analysis
Spatial graph convolutional network
Spatial gene enhancement
Spatially optimal transporting the single cells
Spatial pattern recognition via kernels
Dampened weighted least squares
Spatially resolved transcriptomics with an adaptive graph attention auto-encoder
Copy number variation
Hidden markov random field
Principal component analysis
T-distributed stochastic neighbour embedding
Uniform manifold approximation and projection
Tang F, Barbacioru C, Wang Y, Nordman E, Lee C, Xu N, et al. mRNA-seq whole-transcriptome analysis of a single cell. Nat Methods. 2009;6:377–82.
Longo SK, Guo MG, Ji AL, Khavari PA. Integrating single-cell and spatial transcriptomics to elucidate intercellular tissue dynamics. Nat Rev Genet. 2021;22:627–44.
Chen G, Ning B, Shi T. Single-cell RNA-seq technologies and related computational data analysis. Front Genet. 2019;10:317.
van den Brink SC, Sage F, Vértesy Á, Spanjaard B, Peterson-Maduro J, Baron CS, et al. Single-cell sequencing reveals dissociation-induced gene expression in tissue subpopulations. Nat Methods. 2017;14:935–6.
Ståhl PL, Salmén F, Vickovic S, Lundmark A, Navarro JF, Magnusson J, et al. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science. 2016;353:78–82.
Ortiz C, Navarro JF, Jurek A, Märtin A, Lundeberg J, Meletis K. Molecular atlas of the adult mouse brain. Sci Adv. 2020;6:eabb446.
Hildebrandt F, Andersson A, Saarenpää S, Larsson L, Van Hul N, Kanatani S, et al. Spatial Transcriptomics to define transcriptional patterns of zonation and structural components in the mouse liver. Nat Commun. 2021;12:7046.
Tavares-Ferreira D, Shiers S, Ray PR, Wangzhou A, Jeevakumar V, Sankaranarayanan I, et al. Spatial transcriptomics of dorsal root ganglia identifies molecular signatures of human nociceptors. Sci Transl Med. 2022;14:eabj8186.
Maynard KR, Collado-Torres L, Weber LM, Uytingco C, Barry BK, Williams SR, et al. Transcriptome-scale spatial gene expression in the human dorsolateral prefrontal cortex. Nat Neurosci. 2021;24:425–36.
Asp M, Giacomello S, Larsson L, Wu C, Fürth D, Qian X, et al. A spatiotemporal organ-wide gene expression and cell atlas of the developing human heart. Cell. 2019;179:1647-60.e19.
Hou X, Yang Y, Li P, Zeng Z, Hu W, Zhe R, et al. Integrating spatial transcriptomics and single-cell rna-seq reveals the gene expression profling of the human embryonic liver. Front Cell Dev Biol. 2021;9: 652408.
Fawkner-Corbett D, Antanaviciute A, Parikh K, Jagielowicz M, Gerós AS, Gupta T, et al. Spatiotemporal analysis of human intestinal development at single-cell resolution. Cell. 2021;184:810-26.e23.
Chen H, Murray E, Sinha A, Laumas A, Li J, Lesman D, et al. Dissecting mammalian spermatogenesis using spatial transcriptomics. Cell Rep. 2021;37: 109915.
Joseph DB, Henry GH, Malewska A, Reese JC, Mauck RJ, Gahan JC, et al. 5-Alpha reductase inhibitors induce a prostate luminal to club cell transition in human benign prostatic hyperplasia. J Pathol. 2022;256:427–41.
Maniatis S, Äijö T, Vickovic S, Braine C, Kang K, Mollbrink A, et al. Spatiotemporal dynamics of molecular pathology in amyotrophic lateral sclerosis. Science. 2019;364:89–93.
Boyd DF, Allen EK, Randolph AG, Guo XJ, Weng Y, Sanders CJ, et al. Exuberant fibroblast activity compromises lung function via ADAMTS4. Nature. 2020;587:466–71.
Kuppe C, Ramirez Flores RO, Li Z, Hayat S, Levinson RT, Liao X, et al. Spatial multi-omic map of human myocardial infarction. Nature. 2022;608:766–77.
Moncada R, Barkley D, Wagner F, Chiodin M, Devlin JC, Baron M, et al. Integrating microarray-based spatial transcriptomics and single-cell RNA-seq reveals tissue architecture in pancreatic ductal adenocarcinomas. Nat Biotechnol. 2020;38:333–42.
Ji AL, Rubin AJ, Thrane K, Jiang S, Reynolds DL, Meyers RM, et al. Multimodal analysis of composition and spatial architecture in human squamous cell carcinoma. Cell. 2020;182:497-514.e22.
Andersson A, Larsson L, Stenbeck L, Salmén F, Ehinger A, Wu SZ, et al. Spatial deconvolution of HER2-positive breast cancer delineates tumor-associated cell type interactions. Nat Commun. 2021;12:6012.
Thrane K, Eriksson H, Maaskola J, Hansson J, Lundeberg J. Spatially resolved transcriptomics enables dissection of genetic heterogeneity in stage III cutaneous malignant melanoma. Cancer Res. 2018;78:5970–9.
Satija R, Farrell JA, Gennert D, Schier AF, Regev A. Spatial reconstruction of single-cell gene expression data. Nat Biotechnol. 2015;33:495–502.
Biancalani T, Scalia G, Buffoni L, Avasthi R, Lu Z, Sanger A, et al. Deep learning and alignment of spatially resolved single-cell transcriptomes with Tangram. Nat Methods. 2021;18:1352–62.
Marx V. Method of the Year: spatially resolved transcriptomics. Nat Methods. 2021;18:9–14.
Emmert-Buck MR, Bonner RF, Smith PD, Chuaqui RF, Zhuang Z, Goldstein SR, et al. Laser capture microdissection. Science. 1996;274:998–1001.
Chen Y, Qian W, Lin L, Cai L, Yin K, Jiang S, et al. Mapping gene expression in the spatial dimension. Small Methods. 2021;5: e2100722.
Junker JP, Noël ES, Guryev V, Peterson KA, Shah G, Huisken J, et al. Genome-wide RNA Tomography in the zebrafish embryo. Cell. 2014;159:662–75.
Chen J, Suo S, Tam PP, Han JJ, Peng G, Jing N. Spatial transcriptomic analysis of cryosectioned tissue samples with Geo-seq. Nat Protoc. 2017;12:566–80.
Femino AM, Fay FS, Fogarty K, Singer RH. Visualization of single RNA transcripts in situ. Science. 1998;280:585–90.
Raj A, van den Bogaard P, Rifkin SA, van Oudenaarden A, Tyagi S. Imaging individual mRNA molecules using multiple singly labeled probes. Nat Methods. 2008;5:877–9.
Khong A, Matheny T, Jain S, Mitchell SF, Wheeler JR, Parker R. The stress granule transcriptome reveals principles of mRNA accumulation in stress granules. Mol Cell. 2017;68:808-20.e5.
Codeluppi S, Borm LE, Zeisel A, La Manno G, van Lunteren JA, Svensson CI, et al. Spatial organization of the somatosensory cortex revealed by osmFISH. Nat Methods. 2018;15:932–5.
Lewis SM, Asselin-Labat ML, Nguyen Q, Berthelet J, Tan X, Wimmer VC, et al. Spatial omics and multiplexed imaging to explore cancer biology. Nat Methods. 2021;18:997–1012.
Lubeck E, Coskun AF, Zhiyentayev T, Ahmad M, Cai L. Single-cell in situ RNA profiling by sequential hybridization. Nat Methods. 2014;11:360–1.
Chen KH, Boettiger AN, Moffitt JR, Wang S, Zhuang X. RNA imaging. Spatially resolved, highly multiplexed RNA profiling in single cells. Science. 2015. https://doi.org/10.1126/science.aaa6090.
Xia C, Fan J, Emanuel G, Hao J, Zhuang X. Spatial transcriptome profiling by MERFISH reveals subcellular RNA compartmentalization and cell cycle-dependent gene expression. Proc Natl Acad Sci U S A. 2019;116:19490–9.
Moses L, Pachter L. Museum of spatial transcriptomics. Nat Methods. 2022;19:534–46.
Ke R, Mignardi M, Pacureanu A, Svedlund J, Botling J, Wählby C, et al. In situ sequencing for RNA analysis in preserved tissue and cells. Nat Methods. 2013;10:857–60.
Lee JH, Daugharthy ER, Scheiman J, Kalhor R, Yang JL, Ferrante TC, et al. Highly multiplexed subcellular RNA sequencing in situ. Science. 2014;343:1360–3.
Alon S, Goodwin DR, Sinha A, Wassie AT, Chen F, Daugharthy ER, et al. Expansion sequencing: spatially precise in situ transcriptomics in intact biological systems. 2021. Science. https://doi.org/10.1126/science.aax2656.
Moffitt JR, Lundberg E, Heyn H. The emerging landscape of spatial profiling technologies. Nat Rev Genet. 2022. https://doi.org/10.1038/s41576-022-00515-3.
Wang X, Allen WE, Wright MA, Sylwestrak EL, Samusik N, Vesuna S, et al. Three-dimensional intact-tissue sequencing of single-cell transcriptional states. 2018. Science. https://doi.org/10.1126/science.aat5691.
Rodriques SG, Stickels RR, Goeva A, Martin CA, Murray E, Vanderburg CR, et al. Slide-seq: A scalable technology for measuring genome-wide expression at high spatial resolution. Science. 2019;363:1463–7.
Stickels RR, Murray E, Kumar P, Li J, Marshall JL, Di Bella DJ, et al. Highly sensitive spatial transcriptomics at near-cellular resolution with Slide-seqV2. Nat Biotechnol. 2021;39:313–9.
Vickovic S, Eraslan G, Salmén F, Klughammer J, Stenbeck L, Schapiro D, et al. High-definition spatial transcriptomics for in situ tissue profiling. Nat Methods. 2019;16:987–90.
Cho CS, Xi J, Si Y, Park SR, Hsu JE, Kim M, et al. Microscopic examination of spatial transcriptome using Seq-Scope. Cell. 2021;184:3559-72.e22.
Elhanani O, Ben-Uri R, Keren L. Spatial profiling technologies illuminate the tumor microenvironment. Cancer Cell. 2023;41:404–20.
Wang Y, Song B, Wang S, Chen M, Xie Y, Xiao G, et al. Sprod for de-noising spatially resolved transcriptomics data based on position and image information. Nat Methods. 2022;19:950–8.
Liu Y, Xun Z, Ma K, Liang S, Li X, Zhou S, et al. Identification of a tumour immune barrier in the HCC microenvironment that determines the efficacy of immunotherapy. J Hepatol. 2023;78:770–82.
Galeano Niño JL, Wu H, LaCourse KD, Kempchinsky AG, Baryiames A, Barber B, et al. Effect of the intratumoral microbiota on spatial and cellular heterogeneity in cancer. Nature. 2022;611:810–7.
Wong-Rolle A, Dong Q, Zhu Y, Divakar P, Hor JL, Kedei N, et al. Spatial meta-transcriptomics reveal associations of intratumor bacteria burden with lung cancer cells showing a distinct oncogenic signature. J Immunother Cancer. 2022. https://doi.org/10.1136/jitc-2022-004698.
Bergenstråhle J, Larsson L, Lundeberg J. Seamless integration of image and molecular analysis for spatial transcriptomics workflows. BMC Genomics. 2020;21:482.
Zhao E, Stone MR, Ren X, Guenthoer J, Smythe KS, Pulliam T, et al. Spatial transcriptomics at subspot resolution with BayesSpace. Nat Biotechnol. 2021;39:1375–84.
Hu J, Li X, Coleman K, Schroeder A, Ma N, Irwin DJ, et al. SpaGCN: Integrating gene expression, spatial location and histology to identify spatial domains and spatially variable genes by graph convolutional network. Nat Methods. 2021;18:1342–51.
Svensson V, Teichmann SA, Stegle O. SpatialDE: identification of spatially variable genes. Nat Methods. 2018;15:343–6.
Edsgärd D, Johnsson P, Sandberg R. Identification of spatial expression trends in single-cell gene expression data. Nat Methods. 2018;15:339–42.
Sun S, Zhu J, Zhou X. Statistical analysis of spatial expression patterns for spatially resolved transcriptomic studies. Nat Methods. 2020;17:193–200.
Ma Y, Zhou X. Spatially informed cell-type deconvolution for spatial transcriptomics. Nat Biotechnol. 2022. https://doi.org/10.1038/s41587-022-01273-7.
Elosua-Bayes M, Nieto P, Mereu E, Gut I, Heyn H. SPOTlight: seeded NMF regression to deconvolute spatial transcriptomics spots with single-cell transcriptomes. Nucleic Acids Res. 2021;49: e50.
Cable DM, Murray E, Zou LS, Goeva A, Macosko EZ, Chen F, et al. Robust decomposition of cell type mixtures in spatial transcriptomics. Nat Biotechnol. 2022;40:517–26.
Dong R, Yuan GC. SpatialDWLS: accurate deconvolution of spatial transcriptomic data. Genome Biol. 2021;22:145.
Abdelaal T, Mourragui S, Mahfouz A, Reinders MJT. SpaGE: spatial gene enhancement using scRNA-seq. Nucleic Acids Res. 2020;48: e107.
Yuan Y, Bar-Joseph Z. GCNG: graph convolutional networks for inferring gene interaction from spatial transcriptomics data. Genome Biol. 2020;21:300.
Cang Z, Nie Q. Inferring spatial and signaling relationships between cells from single cell transcriptomic data. Nat Commun. 2020;11:2084.
Patel AP, Tirosh I, Trombetta JJ, Shalek AK, Gillespie SM, Wakimoto H, et al. Single-cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. Science. 2014;344:1396–401.
Erickson A, He M, Berglund E, Marklund M, Mirzazadeh R, Schultz N, et al. Spatially resolved clonal copy number alterations in benign and malignant tissue. Nature. 2022;608:360–7.
Kueckelhaus J, von Ehr J, Ravi VM, Will P, Joseph K, Beck J, et al. Inferring spatially transient gene expression pattern from spatial transcriptomic studies. BioRxiv. 2020. https://doi.org/10.1101/2020.10.20.346544v1.
Moehlin J, Mollet B, Colombo BM, Mendoza-Parra MA. Inferring biologically relevant molecular tissue substructures by agglomerative clustering of digitized spatial transcriptomes with multilayer. Cell Syst. 2021;12:694-705.e3.
Pham D, Tan X, Xu J, Grice LF, Lam PY, Raghubar A, et al. stLearn: integrating spatial location, tissue morphology and gene expression to find cell types, cell-cell interactions and spatial trajectories within undissociated tissues. BioRxiv. 2020. https://doi.org/10.1101/2020.05.31.125658v1.
Wei R, He S, Bai S, Sei E, Hu M, Thompson A, et al. Spatial charting of single-cell transcriptomes in tissues. Nat Biotechnol. 2022. https://doi.org/10.1038/s41587-022-01233-1.
Cui Zhou D, Jayasinghe RG, Chen S, Herndon JM, Iglesia MD, Navale P, et al. Spatially restricted drivers and transitional cell populations cooperate with the microenvironment in untreated and chemo-resistant pancreatic cancer. Nat Genet. 2022;54:1390–405.
Lovatt D, Ruble BK, Lee J, Dueck H, Kim TK, Fisher S, et al. Transcriptome in vivo analysis (TIVA) of spatially defined single cells in live tissue. Nat Methods. 2014;11:190–6.
Shah S, Lubeck E, Schwarzkopf M, He TF, Greenbaum A, Sohn CH, et al. Single-molecule RNA detection at depth by hybridization chain reaction and tissue hydrogel embedding and clearing. Development. 2016;143:2862–7.
Medaglia C, Giladi A, Stoler-Barak L, De Giovanni M, Salame TM, Biram A, et al. Spatial reconstruction of immune niches by combining photoactivatable reporters and scRNA-seq. Science. 2017;358:1622–6.
Chen X, Sun YC, Church GM, Lee JH, Zador AM. Efficient in situ barcode sequencing using padlock probe-based BaristaSeq. Nucleic Acids Res. 2018;46: e22.
Boisset JC, Vivié J, Grün D, Muraro MJ, Lyubimova A, van Oudenaarden A. Mapping the physical network of cellular interactions. Nat Methods. 2018;15:547–53.
Eng CL, Lawson M, Zhu Q, Dries R, Koulena N, Takei Y, et al. Transcriptome-scale super-resolved imaging in tissues by RNA seqFISH. Nature. 2019;568:235–9.
Toki MI, Merritt CR, Wong PF, Smithy JW, Kluger HM, Syrigos KN, et al. High-plex predictive marker discovery for melanoma immunotherapy-treated patients using digital spatial profiling. Clin Cancer Res. 2019;25:5503–12.
Weinstein JA, Regev A, Zhang F. DNA microscopy: optics-free spatio-genetic imaging by a stand-alone chemical reaction. Cell. 2019;178:229-41.e16.
Fazal FM, Han S, Parker KR, Kaewsapsak P, Xu J, Boettiger AN, et al. Atlas of subcellular RNA localization revealed by APEX-Seq. Cell. 2019;178:473-90.e26.
Hu KH, Eichorst JP, McGinnis CS, Patterson DM, Chow ED, Kersten K, et al. ZipSeq: barcoding for real-time mapping of single cell transcriptomes. Nat Methods. 2020;17:833–43.
Liu Y, Yang M, Deng Y, Su G, Enninful A, Guo CC, et al. High-spatial-resolution multi-omics sequencing via deterministic barcoding in tissue. Cell. 2020;183:1665-81.e18.
Lee Y, Bogdanoff D, Wang Y, Hartoularos GC, Woo JM, Mowery CT, et al. XYZeq: spatially resolved single-cell RNA sequencing reveals expression heterogeneity in the tumor microenvironment. Sci Adv. 2021. https://doi.org/10.1126/sciadv.abg4755.
Srivatsan SR, Regier MC, Barkan E, Franks JM, Packer JS, Grosjean P, et al. Embryo-scale, single-cell spatial transcriptomics. Science. 2021;373:111–7.
Chen A, Liao S, Cheng M, Ma K, Wu L, Lai Y, et al. Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA nanoball-patterned arrays. Cell. 2022;185:1777-92.e21.
Fan Y, Andrusivová Ž, Wu Y, Chai C, Larsson L, He M, et al. Expansion spatial transcriptomics. BioRxiv. 2022;19:534.
Bergenstråhle J, Bergenstråhle L, Lundeberg J. SpatialCPie: an R/Bioconductor package for spatial transcriptomics cluster evaluation. BMC Bioinformatics. 2020;21:161.
Andersson A, Bergenstråhle J, Asp M, Bergenstråhle L, Jurek A, Fernández Navarro J, et al. Single-cell and spatial transcriptomics enables probabilistic inference of cell type topography. Commun Biol. 2020;3:565.
Song Q, Su J. DSTG: deconvoluting spatial transcriptomics data through graph-based artificial intelligence. Brief Bioinform. 2021. https://doi.org/10.1093/bib/bbaa414.
Dries R, Zhu Q, Dong R, Eng CL, Li H, Liu K, et al. Giotto: a toolbox for integrative analysis and visualization of spatial expression data. Genome Biol. 2021;22:78.
Hao M, Hua K, Zhang X. SOMDE: a scalable method for identifying spatially variable genes with self-organizing map. Bioinformatics. 2021;1:701680.
Kleshchevnikov V, Shmatko A, Dann E, Aivazidis A, King HW, Li T, et al. Cell 2location maps fine-grained cell types in spatial transcriptomics. Nat Biotechnol. 2022. https://doi.org/10.1038/s41587-021-01139-4.
Dong K, Zhang S. Deciphering spatial domains from spatially resolved transcriptomics with an adaptive graph attention auto-encoder. Nat Commun. 2022;13:1739.
We appreciate all the members of our research team on Lymphatic System Tumor for the proofreading and valuable feedback. We also thank the support from the School of Medicine, Shanghai Jiao Tong University. Parts of the elements in the figure were drawn after modification and adaptation of pictures from Servier Medical Art by Servier, licensed under a Creative Commons Attribution 3.0 Unported License.
This project was funded by Shanghai Shenkang Hospital Development Center (No. SHDC2020CR2070B).
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Du, J., Yang, YC., An, ZJ. et al. Advances in spatial transcriptomics and related data analysis strategies. J Transl Med 21, 330 (2023). https://doi.org/10.1186/s12967-023-04150-2