Spatial Gene Expression Research Articles

Spatial differences in gene expression across the dorsal raphe nucleus in a model of early Alzheimer's disease.

Persons with Alzheimer's disease (AD) present with changes in mood, sleep, and arousal that may precede the clinical manifestation of cognitive decline. These early symptoms can be driven by changes in the serotonergic (5-HT) nuclei of the brainstem, particularly the dorsal raphe nucleus (DRN). It is unclear why all 5-HT neurons do not simultaneously develop AD pathology that progresses at the same rate. We sought to identify any underlying genetic components associated with susceptibility or resistance of 5-HT neurons to AD pathology. The Visium Spatial Gene Expression platform was used to identify transcriptomic changes across the DRN in a preclinical model of early AD, human tau-overexpressing mice (htau mice). We further used RNAscope and immunohistochemical assessment to validate findings of primary interest. We find that the DRN of htau mice differentially expresses AD-related genes, including those related to kinase binding, ion channel activity, ligand-receptor interactions, and regulation of serine/threonine kinases. We further find that computational sub-clustering of the DRN is consistent with previous circuitry-driven characterizations, allowing for spatial bounding of distinct subregions within the DRN. Of these, we find the dorsolateral DRN is preferentially impacted by 5-HT neuron loss and development of tau pathology, which coincides with increased expression of the long noncoding RNA Map2k3os. Map2k3os may serve regulatory roles relevant for tau phosphorylation and warrants further investigation to characterize its interactions. Overall, this report demonstrates the power of large-scale spatial transcriptomics technologies, while underscoring the need for convergent-data validation to overcome their limitations.

Regional neural functional efficiency across schizophrenia, bipolar disorder, and major depressive disorder: a transdiagnostic resting-state fMRI study

Abstract Background Major psychiatric disorders (MPDs) are delineated by distinct clinical features. However, overlapping symptoms and transdiagnostic effectiveness of medications have challenged the traditional diagnostic categorisation. We investigate if there are shared and illness-specific disruptions in the regional functional efficiency (RFE) of the brain across these disorders. Methods We included 364 participants (118 schizophrenia [SCZ], 80 bipolar disorder [BD], 91 major depressive disorder [MDD], and 75 healthy controls [HCs]). Resting-state fMRI was used to caclulate the RFE based on the static amplitude of low-frequency fluctuation, regional homogeneity, and degree centrality and corresponding dynamic measures indicating variability over time. We used principal component analysis to obtain static and dynamic RFE values. We conducted functional and genetic annotation and enrichment analysis based on abnormal RFE profiles. Results SCZ showed higher static RFE in the cortico-striatal regions and excessive variability in the cortico-limbic regions. SCZ and MDD shared lower static RFE with higher dynamic RFE in sensorimotor regions than BD and HCs. We observed association between static RFE abnormalities with reward and sensorimotor functions and dynamic RFE abnormalities with sensorimotor functions. Differential spatial expression of genes related to glutamatergic synapse and calcium/cAMP signaling was more likely in the regions with aberrant RFE. Conclusions SCZ shares more regions with disrupted functional integrity, especially in sensorimotor regions, with MDD rather than BD. The neural patterns of these transdiagnostic changes appear to be potentially driven by gene expression variations relating to glutamatergic synapses and calcium/cAMP signaling. The aberrant sensorimotor, cortico-striatal, and cortico-limbic integrity may collectively underlie neurobiological mechanisms of MPDs.

Advanced single-cell and spatial analysis with high-multiplex characterization of circulating tumor cells and tumor tissue in prostate cancer: Unveiling resistance mechanisms with the CoDuCo in situ assay

BackgroundMetastatic prostate cancer is a highly heterogeneous and dynamic disease and practicable tools for patient stratification and resistance monitoring are urgently needed. Liquid biopsy analysis of circulating tumor cells (CTCs) and circulating tumor DNA are promising, however, comprehensive testing is essential due to diverse mechanisms of resistance. Previously, we demonstrated the utility of mRNA-based in situ padlock probe hybridization for characterizing CTCs.MethodsWe have developed a novel combinatorial dual-color (CoDuCo) assay for in situ mRNA detection, with enhanced multiplexing capacity, enabling the simultaneous analysis of up to 15 distinct markers. This approach was applied to CTCs, corresponding tumor tissue, cancer cell lines, and peripheral blood mononuclear cells for single-cell and spatial gene expression analysis. Using supervised machine learning, we trained a random forest classifier to identify CTCs. Image analysis and visualization of results was performed using open-source Python libraries, CellProfiler, and TissUUmaps.ResultsOur study presents data from multiple prostate cancer patients, demonstrating the CoDuCo assay’s ability to visualize diverse resistance mechanisms, such as neuroendocrine differentiation markers (SYP, CHGA, NCAM1) and AR-V7 expression. In addition, druggable targets and predictive markers (PSMA, DLL3, SLFN11) were detected in CTCs and formalin-fixed, paraffin-embedded tissue. The machine learning-based CTC classification achieved high performance, with a recall of 0.76 and a specificity of 0.99.ConclusionsThe combination of high multiplex capacity and microscopy-based single-cell analysis is a unique and powerful feature of the CoDuCo in situ assay. This synergy enables the simultaneous identification and characterization of CTCs with epithelial, epithelial-mesenchymal, and neuroendocrine phenotypes, the detection of CTC clusters, the visualization of CTC heterogeneity, as well as the spatial investigation of tumor tissue. This assay holds significant potential as a tool for monitoring dynamic molecular changes associated with drug response and resistance in prostate cancer.

Digital profiling of gene expression from histology images with linearized attention

Cancer is a heterogeneous disease requiring costly genetic profiling for better understanding and management. Recent advances in deep learning have enabled cost-effective predictions of genetic alterations from whole slide images (WSIs). While transformers have driven significant progress in non-medical domains, their application to WSIs lags behind due to high model complexity and limited dataset sizes. Here, we introduce SEQUOIA, a linearized transformer model that predicts cancer transcriptomic profiles from WSIs. SEQUOIA is developed using 7584 tumor samples across 16 cancer types, with its generalization capacity validated on two independent cohorts comprising 1368 tumors. Accurately predicted genes are associated with key cancer processes, including inflammatory response, cell cycles and metabolism. Further, we demonstrate the value of SEQUOIA in stratifying the risk of breast cancer recurrence and in resolving spatial gene expression at loco-regional levels. SEQUOIA hence deciphers clinically relevant information from WSIs, opening avenues for personalized cancer management.

The E3 ubiquitin ligase RNF220 maintains hindbrain Hox expression patterns through regulation of WDR5 stability.

The spatial and temporal linear expression of Hox genes establishes a regional Hox code, which is crucial for the antero-posterior (A-P) patterning, segmentation, and neuronal circuit development of the hindbrain. RNF220, an E3 ubiquitin ligase, is widely involved in neural development via targeting of multiple substrates. Here, we found that the expression of Hox genes in the pons was markedly up-regulated at the late developmental stage (post-embryonic day E15.5) in Rnf220-/- and Rnf220+/- mouse embryos. Single-nucleus RNA sequencing (RNA-seq) analysis revealed different Hox de-repression profiles in different groups of neurons, including the pontine nuclei (PN). The Hox pattern was disrupted and the neural circuits were affected in the PN of Rnf220+/- mice. We showed that this phenomenon was mediated by WDR5, a key component of the TrxG complex, which can be polyubiquitinated and degraded by RNF220. Intrauterine injection of WDR5 inhibitor (WDR5-IN-4) and genetic ablation of Wdr5 in Rnf220+/- mice largely recovered the de-repressed Hox expression pattern in the hindbrain. In P19 embryonal carcinoma cells, the retinoic acid-induced Hox expression was further stimulated by Rnf220 knockdown, which can also be rescued by Wdr5 knockdown. In short, our data suggest a new role of RNF220/WDR5 in Hox pattern maintenance and pons development in mice.

The E3 ubiquitin ligase RNF220 maintains hindbrain Hox expression patterns through regulation of WDR5 stability

The spatial and temporal linear expression of Hox genes establishes a regional Hox code, which is crucial for the antero-posterior (A-P) patterning, segmentation, and neuronal circuit development of the hindbrain. RNF220, an E3 ubiquitin ligase, is widely involved in neural development via targeting of multiple substrates. Here, we found that the expression of Hox genes in the pons was markedly up-regulated at the late developmental stage (post-embryonic day E15.5) in Rnf220-/- and Rnf220+/- mouse embryos. Single-nucleus RNA sequencing (RNA-seq) analysis revealed different Hox de-repression profiles in different groups of neurons, including the pontine nuclei (PN). The Hox pattern was disrupted and the neural circuits were affected in the PN of Rnf220+/- mice. We showed that this phenomenon was mediated by WDR5, a key component of the TrxG complex, which can be polyubiquitinated and degraded by RNF220. Intrauterine injection of WDR5 inhibitor (WDR5-IN-4) and genetic ablation of Wdr5 in Rnf220+/- mice largely recovered the de-repressed Hox expression pattern in the hindbrain. In P19 embryonal carcinoma cells, the retinoic acid-induced Hox expression was further stimulated by Rnf220 knockdown, which can also be rescued by Wdr5 knockdown. In short, our data suggest a new role of RNF220/WDR5 in Hox pattern maintenance and pons development in mice.

ANGI-08. SPATIAL TRANSCRIPTOMIC COMPARISONS OF GLIOBLASTOMA TISSUE REVEAL UPREGULATION OF NEURODEVELOPMENTAL PATHWAYS AND SYNAPTIC GENES IN INFILTRATING TUMOR CELLS

Abstract Glioblastoma remains one of the deadliest brain malignancies. First-line therapy consists of maximal surgical tumor resection, chemotherapy, and radiotherapy. Malignant cells escape surgical resection by migrating into the surrounding healthy brain tissue, where they give rise to the recurrent tumor. Gene expression profiling allows glioblastoma tumor cores to be classified into mesenchymal, proneural, and classical subtypes, each with distinct genetic alterations and cellular compositions. In contrast, the adjacent brain parenchyma where infiltrating malignant cells escape surgical resection is less characterized in patients. Using single-cell and multicellular resolution spatial transcriptomics on tissues from both the tumor core and infiltrated brain tissue (n = 11), we compared the transcriptional profiles of malignant cells in these regions. Malignant cells near regions with microvascular proliferation showed increased expression of genes related to vascular homeostasis. Within tumor cores, proneural and mesenchymal subtypes exhibited distinct spatial gene expression patterns, although these differences were reduced in the infiltrated brain tissue, apart from gene expression due to chromosomal alterations specific to the proneural subtype. We identified two transcriptional patterns of brain infiltration, with the dominant pattern showing a transition from mesenchymal to proneural states. This transition was accompanied by increased expression of genes linked to neurodevelopmental pathways and glial cell differentiation. Additionally, one gene module upregulated in infiltrating malignant cells was predictive of poor patient survival and enriched for genes associated with differentiated neuronal cells. Together, our findings provide an updated view of the spatial landscape of glioblastomas and infiltrated brain tissue, furthering our understanding of the malignant cells that invade healthy brain. This insight offers new avenues for targeted therapy after surgical resection of the primary tumor.

Ecdysteroid-dependent molting in tardigrades.

Although molting is a defining feature of the most species-rich animal taxa-the Ecdysozoa, including arthropods, tardigrades, nematodes, and others1,2-its evolutionary background remains enigmatic. In pancrustaceans, such as insects and decapods, molting is regulated by the ecdysteroid (Ecd) hormone and its downstream cascade (Figure1A, see also the text).3,4,5 However, whether Ecd-dependent molting predates the emergence of the arthropods and represents an ancestral machinery in ecdysozoans remains unclear. For example, involvement of the Ecd hormone in molting regulation has been suggested only in some parasitic nematodes outside of arthropods,6,7 and insect Ecd synthesis and receptor genes are lacking in some ecysozoan lineages (FigureS1A).8,9,10 In this study, we investigated the role of Ecd in the molting process of the tardigrade Hypsibius exemplaris. We show that the endogenous Ecd level periodically increases during the molting cycle of H.exemplaris. The pulse treatment with exogenous Ecd induced molting, whereas an antagonist of the Ecd receptor suppressed the molting. Our spatial and temporal gene expression analysis revealed the putative regulatory organs and Ecd downstream cascades. We demonstrate that tardigrade molting is regulated by the Ecd hormone, supporting the ancestry of Ecd-dependent molting in panarthropods. Furthermore, we were able to identify the putative neural center of molting regulation in tardigrades. This region may be homologous to the neural center in the protocerebrum of pancrustaceans and represent an ancestral state of panarthropods. Together, our results suggest that Ecd-dependent molting evolved in the early-late Ediacaran, 22-76 million years earlier than previously suggested.11.

SpaInGNN: Enhanced clustering and integration of spatial transcriptomics based on refined graph neural networks

Recent developments in spatial transcriptomics (ST) technology have markedly enhanced the proposed capacity to comprehensively characterize gene expression patterns within tissue microenvironments while crucially preserving spatial context. However, the identification of spatial domains at the single-cell level remains a significant challenge in elucidating biological processes. To address this, SpaInGNN was developed, a sophisticated graph neural network (GNN) framework that accurately delineates spatial domains by integrating spatial location data, histological information, and gene expression profiles into low-dimensional latent embeddings. Additionally, to fully leverage spatial coordinate data, spatial integration using graph neural network (SpaInGNN) refines the graph constructed for spatial locations by incorporating both tissue image distance and Euclidean distance, following a pre-clustering of gene expression profiles. This refined graph is then embedded using a self-supervised GNN, which minimizes self-reconfiguration loss. By applying SpaInGNN to refined graphs across multiple consecutive tissue slices, this study mitigates the impact of batch effects in data analysis. The proposed method demonstrates substantial improvements in the accuracy of spatial domain recognition, providing a more faithful representation of the tissue organization in both mouse olfactory bulb and human lateral prefrontal cortex samples

ITGA5+ synovial fibroblasts orchestrate proinflammatory niche formation by remodelling the local immune microenvironment in rheumatoid arthritis.

To investigate the phenotypic heterogeneity of tissue-resident synovial fibroblasts and their role in inflammatory response in rheumatoid arthritis (RA). We used single-cell and spatial transcriptomics to profile synovial cells and spatial gene expressions of synovial tissues to identify phenotypic changes in patients with osteoarthritis, RA in sustained remission and active state. Immunohistology, multiplex immunofluorescence and flow cytometry were used to identify synovial fibroblasts subsets. Deconvolution methods further validated our findings in two cohorts (PEAC and R4RA) with treatment response. Cell coculture was used to access the potential cell-cell interactions. Adoptive transfer of synovial cells in collagen-induced arthritis (CIA) mice and bulk RNA sequencing of synovial joints further validate the cellular functions. We identified a novel tissue-remodelling CD45-CD31-PDPN+ITGA5+ synovial fibroblast population with unique transcriptome of POSTN, COL3A1, CCL5 and TGFB1, and enriched in immunoregulatory pathways. This subset was upregulated in active and lympho-myeloid type of RA, associated with an increased risk of multidrug resistance. Transforming growth factor (TGF)-β1 might participate in the differentiation of this subset. Moreover, ITGA5+ synovial fibroblasts might occur in early stage of inflammation and induce the differentiation of CXCL13hiPD-1hi peripheral helper T cells (TPHs) from naïve CD4+ T cells, by secreting TGF-β1. Intra-articular injection of ITGA5+ synovial fibroblasts exacerbates RA development and upregulates TPHs in CIA mice. We demonstrate that ITGA5+ synovial fibroblasts might regulate the RA progression by inducing the differentiation of CXCL13hiPD-1hi TPHs and remodelling the proinflammatory microenvironments. Therapeutic modulation of this subpopulation could therefore be a potential treatment strategy for RA.

Role of inflammation and macrophages in the formation of the atrial cardiomyopathy associated with obesity

Abstract Introduction The atrial fibrillation (AF), the most frequent cardiac arrhythmia in clinical practice is often associated with an atrial cardiomyopathy (ACM) which is favoured by several clinical conditions including metabolic diseases such as obesity. Recently, the epicardial adipose tissue (EAT) has emerged as an important component of ACM notably in the context of metabolic diseases. Epicardial progenitor cells (EPC) are the main source of EAT. The replacement of EAT by fibrosis is arrhythmogenic and involved inflammation but not yet characterized. The hypothesis tested here is that macrophages are central players for both the recruitment of EPC and replacement of EAT by fibrosis. Purpose-Aims First to characterize subpopulation of macrophages involves in the progression of ACM. Second, to study how the inflammatory environment regulates EPC differentiation. Method A spatial gene expression was used to analyze human samples (n=4) (Agreement, CODECOH, Declaration or authorization of activities involving the preservation and preparation of human body parts for scientific purposes DC-2020-4183). A well-characterized mouse model of atrial cardiomyopathy associated with obesity induced by high fat diet (HFD) was used to characterize macrophages associated with progression of ACM by using immunolabeling together with single nuclei RNA sequencing approaches. ACM was characterized by using echocardiography and atrial burst pacing to evaluate determine AF vulnerability. Transgenic RGB-MAC mice with macrophages tagged with McApple-Csf1R and ECGFP-CX3CR1 or with a knock-out for CCR2 were used to follow infiltration of myeloid cells in the atrial at the different stages of the ACM. Bi-photon microscopy was used to visualize cell subpopulations in the distinct area of the atrial wall. In some experiments EPC and macrophages were cocultured with transwell. Results Macrophages were detected in subepicardium of human atrial samples in area of fibro-fatty infiltrates. After 2 months of HFD, monocytes were infiltrated atria and macrophages were overexpressed in atria from 1 to 4 months of HFD. Using Bi-photon microscopy it was possible to visualise resident ECGFP-CX3CR1+ macrophages at the contact of EAT whereas McApple-Csfr1r+ macrophages were at the junction with fibrosis (figure 1). Preliminary data generated by single nuclei RNA sequencing showed clusters of different subpopulations of macrophages at different time point of the ACM (figure 2). We will present results on the effects of suppression of the CCR2 subpopulation of macrophages on characteristics of the ACM. Conclusion Distinct populations of macrophages i.e resident and monocyte-derived macrophages, participate to the distinct stages of the ACM caused by obesity in mice.Myeloid cells infiltration in epicardiumCCR2+cells infiltration in EAT-fibrosis

Graph attention automatic encoder based on contrastive learning for domain recognition of spatial transcriptomics

Spatial transcriptomics is an emerging technology that enables the profiling of gene expression in tissues while preserving spatial location information. This innovative approach is anticipated to provide a comprehensive understanding of the spatial distribution of different cells within tissues and facilitate in-depth analysis of tissue structure. To accurately recognize spatial domains from spatial transcriptomics, we have introduced a generalized deep learning method called GAAEST (Graph Attention-based Autoencoder for Spatial Transcriptomics). Our proposed approach effectively integrates both spatial location information and gene expression data from spatial transcriptomics. Specifically, it leverages spatial location details to construct a neighborhood graph and employs a graph attention network-based encoder to embed gene expression information into a spatially informed space. At the same time, to further optimize the learned potential embedding, self-supervised contrastive learning is introduced to capture spatial information at three levels: local, global and contextual feature of spots. Finally, the decoder reconstructs gene expressions, which are then clustered to identify spatial domains with similar expression patterns and spatial proximity. Based on our experiments conducted on multiple datasets, GAAEST consistently outperforms existing state-of-the-art methods. The proposed GAAEST demonstrates excellent capabilities in spatial domain recognition, positioning it as an ideal tool for advancing spatial transcriptomics research.

Single Cell Sequencing Provides Clues about the Developmental Genetic Basis of Evolutionary Adaptations in Syngnathid Fishes.

Seahorses, pipefishes, and seadragons are fishes from the family Syngnathidae that have evolved extraordinary traits including male pregnancy, elongated snouts, loss of teeth, and dermal bony armor. The developmental genetic and cellular changes that led to the evolution of these traits are largely unknown. Recent syngnathid genome assemblies revealed suggestive gene content differences and provide the opportunity for detailed genetic analyses. We created a single cell RNA sequencing atlas of Gulf pipefish embryos to understand the developmental basis of four traits: derived head shape, toothlessness, dermal armor, and male pregnancy. We completed marker gene analyses, built genetic networks, and examined spatial expression of select genes. We identified osteochondrogenic mesenchymal cells in the elongating face that express regulatory genes bmp4, sfrp1a , and prdm16 . We found no evidence for tooth primordia cells, and we observed re-deployment of osteoblast genetic networks in developing dermal armor. Finally, we found that epidermal cells expressed nutrient processing and environmental sensing genes, potentially relevant for the brooding environment. The examined pipefish evolutionary innovations are composed of recognizable cell types, suggesting derived features originate from changes within existing gene networks. Future work addressing syngnathid gene networks across multiple stages and species is essential for understanding how their novelties evolved.

Dual spatial host-bacterial gene expression in Mycobacterium abscessus respiratory infections

Co-localization of spatial transcriptome information of host and pathogen can revolutionize our understanding of microbial pathogenesis. Here, we aimed to demonstrate that customized bacterial probes can be successfully used to identify host-pathogen interactions in formalin-fixed-paraffin-embedded (FFPE) tissues by probe-based spatial transcriptomics technology. We analyzed the spatial gene expression of bacterial transcripts with the host transcriptomic profile in murine lung tissue chronically infected with Mycobacterium abscessus embedded in agar beads. Customized mycobacterial probes were designed for the constitutively expressed rpoB gene (an RNA polymerase β subunit) and the virulence factor precursor lsr2, modulated by oxidative stress. We found a correlation between the rpoB expression, bacterial abundance in the airways, and an increased expression of lsr2 virulence factor in lung tissue with high oxidative stress. Overall, we demonstrate the potential of dual bacterial and host gene expression assay in FFPE tissues, paving the way for the simultaneous detection of host and bacterial transcriptomes in pathological tissues.

HCC spatial transcriptomic profiling reveals significant and potentially targetable cancer-endothelial interactions.

HCC is a highly vascular tumor, and many effective drug regimens target the tumor blood vessels. Prior bulk HCC subtyping data used bulk transcriptomes, which contained a mixture of parenchymal and stromal contributions. We utilized computational deconvolution and cell-cell interaction analyses to cell type-specific (tumor-enriched and vessel-enriched) spatial transcriptomic data collected from 41 resected HCC tissue specimens. We report that the prior Hoshida bulk transcriptional subtyping schema is driven largely by an endothelial fraction, show an alternative tumor-specific schema has potential prognostic value, and use spatially paired ligand-receptor analyses to identify known and novel (LGALS9 tumor-HAVCR2 vessel) signaling relationships that drive HCC biology in a subtype-specific and potentially targetable manner. Our study leverages spatial gene expression profiling technologies to dissect HCC heterogeneity and identify heterogeneous signaling relationships between cancer cells and their endothelial cells. Future validation and expansion of these findings may validate novel cancer-endothelial cell interactions and related drug targets.

Spatial Transcriptomic Profiling of Human Saphenous Vein Exposed to Ex Vivo Arterial Haemodynamics-Implications for Coronary Artery Bypass Graft Patency and Vein Graft Disease.

Vein graft disease is the process by which saphenous vein grafts, utilised for revascularisation during coronary artery bypass graft surgery, undergo an inflammation-driven intimal hyperplasia and accelerated atherosclerosis process in subsequent years after implantation. The role of the arterial circulation, particularly the haemodynamic properties' impact on graft patency, have been investigated but have not to date been explored in depth at the transcriptomic level. We have undertaken the first-in-man spatial transcriptomic analysis of the long saphenous vein in response to ex vivo acute arterial haemodynamic stimulation, utilising a combination of a custom 3D-printed perfusion bioreactor and the 10X Genomics Visium Spatial Gene Expression technology. We identify a total of 413 significant genes (372 upregulated and 41 downregulated) differentially expressed in response to arterial haemodynamic conditions. These genes were associated with pathways including NFkB, TNF, MAPK, and PI3K/Akt, among others. These are established pathways involved in the initiation of an early pro-inflammatory response, leukocyte activation and adhesion signalling, tissue remodelling, and cellular differentiation. Utilising unsupervised clustering analysis, we have been able to classify subsets of the expression based on cell type and with spatial resolution. These findings allow for further characterisation of the early saphenous vein graft transcriptional landscape during the earliest stage of implantation that contributes to vein graft disease, in particular validation of pathways and druggable targets that could contribute towards the therapeutic inhibition of processes underpinning vein graft disease.

SpaDiT: diffusion transformer for spatial gene expression prediction using scRNA-seq.

The rapid development of spatially resolved transcriptomics (SRT) technologies has provided unprecedented opportunities for exploring the structure of specific organs or tissues. However, these techniques (such as image-based SRT) can achieve single-cell resolution, but can only capture the expression levels of tens to hundreds of genes. Such spatial transcriptomics (ST) data, carrying a large number of undetected genes, have limited its application value. To address the challenge, we develop SpaDiT, a deep learning framework for spatial reconstruction and gene expression prediction using scRNA-seq data. SpaDiT employs scRNA-seq data as an a priori condition and utilizes shared genes between ST and scRNA-seq data as latent representations to construct inputs, thereby facilitating the accurate prediction of gene expression in ST data. SpaDiT enhances the accuracy of spatial gene expression predictions over a variety of spatial transcriptomics datasets. We have demonstrated the effectiveness of SpaDiT by conducting extensive experiments on both seq-based and image-based ST data. We compared SpaDiT with eight highly effective baseline methods and found that our proposed method achieved an 8%-12% improvement in performance across multiple metrics. Source code and all datasets used in this paper are available at https://github.com/wenwenmin/SpaDiT and https://zenodo.org/records/12792074.

Multimodal contrastive learning for spatial gene expression prediction using histology images.

In recent years, the advent of spatial transcriptomics (ST) technology has unlocked unprecedented opportunities for delving into the complexities of gene expression patterns within intricate biological systems. Despite its transformative potential, the prohibitive cost of ST technology remains a significant barrier to its widespread adoption in large-scale studies. An alternative, more cost-effective strategy involves employing artificial intelligence to predict gene expression levels using readily accessible whole-slide images stained with Hematoxylin and Eosin (H&E). However, existing methods have yet to fully capitalize on multimodal information provided by H&E images and ST data with spatial location. In this paper, we propose mclSTExp, a multimodal contrastive learning with Transformer and Densenet-121 encoder for Spatial Transcriptomics Expression prediction. We conceptualize each spot as a "word", integrating its intrinsic features with spatial context through the self-attention mechanism of a Transformer encoder. This integration is further enriched by incorporating image features via contrastive learning, thereby enhancing the predictive capability of our model. We conducted an extensive evaluation of highly variable genes in two breast cancer datasets and a skin squamous cell carcinoma dataset, and the results demonstrate that mclSTExp exhibits superior performance in predicting spatial gene expression. Moreover, mclSTExp has shown promise in interpreting cancer-specific overexpressed genes, elucidating immune-related genes, and identifying specialized spatial domains annotated by pathologists. Our source code is available at https://github.com/shizhiceng/mclSTExp.

Cross-expression analysis reveals patterns of coordinated gene expression in spatial transcriptomics.

Spatial transcriptomics promises to transform our understanding of tissue biology by molecularly profiling individual cells in situ. A fundamental question they allow us to ask is how nearby cells orchestrate their gene expression. To investigate this, we introduce cross-expression, a novel framework for discovering gene pairs that coordinate their expression across neighboring cells. Just as co-expression quantifies synchronized gene expression within the same cells, cross-expression measures coordinated gene expression between spatially adjacent cells, allowing us to understand tissue gene expression programs with single cell resolution. Using this framework, we recover ligand-receptor partners and discover gene combinations marking anatomical regions. More generally, we create cross-expression networks to find gene modules with orchestrated expression patterns. Finally, we provide an efficient R package to facilitate cross-expression analysis, quantify effect sizes, and generate novel visualizations to better understand spatial gene expression programs.

AFSC: A self-supervised augmentation-free spatial clustering method based on contrastive learning for identifying spatial domains

Recent research in spatial transcriptomics allows researchers to analyze gene expression without losing spatial information. Spatial information can assist in cell communication, identification of new cell subtypes, which provides important research methods for multiple fields such as microenvironment interactions and pathological processes of diseases. Identifying spatial domains is an important step in spatial transcriptomics analysis, and improving spatial clustering methods can benefit for identifying spatial domains. In addition to eliminating noise in original gene expression, how to use spatial information to assist clustering has also become a new problem. A variety of calculating methods have been applied to spatial clustering, including contrastive learning methods. However, existing spatial clustering methods based on contrastive learning use data augmentation to generate positive and negative pairs, which will inevitably destroy the biological meaning of the data. We propose a new self-supervised spatial clustering method based on contrastive learning, Augmentation-Free Spatial Clustering (AFSC), which integrates spatial information and gene expression to learn latent representations. We construct a contrastive learning module without negative pairs or data augmentation by designing Teacher and Student Encoder. We also design an unsupervised clustering module to make clustering and contrastive learning be trained together. Experiments on multiple spatial transcriptomics datasets at different resolutions demonstrate that our method performs well in self-supervised spatial clustering tasks. Furthermore, the learned representations can be used for various downstream tasks including visualization and trajectory inference.