Spatial transcriptomics (ST) enables the simultaneous measurement of gene expression and spatial localization within tissue sections, providing unprecedented opportunities to dissect tissue architecture and functional organization. As a relatively new omics technology, bioinformatics has driven much of the innovation in ST. However, within these frameworks, spatial information is often reduced to locations and relationships between molecular profiles, without fully leveraging the wealth of sub-micron morphological detail and histological knowledge available. Advances in computer vision-based artificial intelligence (AI) are opening exciting new avenues beyond conventional bioinformatics approaches by modeling complex histological patterns and linking morphology to molecular states. More excitingly, they bring fresh perspectives to potentially address key limitations of ST, including its high cost, limited clinical applicability, and reliance on 2D analysis of inherently 3D tissues. For instance, models that predict ST directly from histology images enable virtual sequencing, drastically reducing costs while integrating morphological insights from pathology with molecular biomarkers, thus accelerating clinical translation. Moreover, computer vision techniques can reconstruct pixel-aligned 3D tissue models, overcoming the technical barriers of 2D acquisition and advancing 3D spatial omics analytics. In this paper, we present the first systematic survey of computer vision AI models for ST analytics, categorizing approaches across architectures, learning paradigms, tasks, and datasets, and tracing their technological evolution. We highlight key challenges and future directions, offering a panoramic perspective on vision-driven ST and its potential to transform both basic research and clinical practice. The curated collection of vision-driven ST papers is available at https://github.com/hrlblab/computer_vision_spatial_omics.

Spatial transcriptomics (ST) is an emerging technology that enables medical computer vision scientists to automatically interpret the molecular profiles underlying morphological features. Currently, however, most deep learning-based ST analyses are limited to two-dimensional (2D) sections, which can introduce diagnostic errors due to the heterogeneity of pathological tissues across 3D sections. Expanding ST to three-dimensional (3D) volumes is challenging due to the prohibitive costs; a 2D ST acquisition already costs over 50 times more than whole slide imaging (WSI), and a full 3D volume with 10 sections can be an order of magnitude more expensive. To reduce costs, scientists have attempted to predict ST data directly from WSI without performing actual ST acquisition. However, these methods typically yield unsatisfying results. To address this, we introduce a novel problem setting: 3D ST imputation using 3D WSI histology sections combined with a single 2D ST slide. To do so, we present the Anatomy-aware Spatial Imputation Graph Network (ASIGN) for more precise, yet affordable, 3D ST modeling. The ASIGN architecture extends existing 2D spatial relationships into 3D by leveraging cross-layer overlap and similarity-based expansion. Moreover, a multi-level spatial attention graph network integrates features comprehensively across different data sources. We evaluated ASIGN on three public spatial transcriptomics datasets, with experimental results demonstrating that ASIGN achieves state-of-the-art performance on both 2D and 3D scenarios. The code for this paper is publicly available1.

In contrast with human sketching, which pre-conceptualizes outlines and features, conventional sketch retrieval models rely primarily rely on pixel-level processing and feature extraction, limiting their ability to capture early sketch intent. Consequently, these models are susceptible to subjective stroke noise, reducing retrieval accuracy. To address this issue, we propose a novel on-the-fly noise stroke retrieval framework designed to align with human sketch-drawing cognition. The proposed framework introduces two core innovations. (i) A stroke consistency detection module that effectively discriminates and suppresses noise strokes by quantifying the structural similarity between the current stroke and the target image, as well as its alignment with key skeletal components. (ii) An adaptive gated mixture of experts module that dynamically selects and integrates features from multiple expert networks during the early, sparse stages of sketching, thereby capturing relevant information with greater precision. Experimental results across diverse sketch datasets demonstrate that the proposed method effectively identifies and suppresses early noise strokes, significantly enhances sketch retrieval performance, and exhibits strong robustness across varying sketch styles.

Spatial transcriptomics (ST) integrates gene expression data with the spatial organization of cells and their associated histology, offering unprecedented insights into tissue biology. While existing methods incorporate either location-based or histology-informed information, none fully synergize gene expression, histological features, and precise spatial coordinates within a unified framework. Moreover, these methods often exhibit inconsistent performance across diverse datasets and conditions. Here, we introduce stImage, an open-source R package that provides a comprehensive and flexible solution for ST analysis. By generating deep learning-derived histology features and offering 54 integrative strategies, stImage seamlessly combines transcriptional profiles, histology images, and spatial information. We demonstrate stImage's effectiveness across multiple datasets, underscoring its ability to guide users toward the most suitable integration strategy using diagnostic graph. Our results highlight how stImage can optimize ST, consistently improving biological insights and advancing our understanding of tissue architecture. stImage is freely available at https://github.com/YuWang-VUMC/stImage.

Understanding the anatomy of renal pathology is crucial for advancing disease diagnostics, treatment evaluation, and clinical research. The complex kidney system comprises various components across multiple levels, including regions (cortex, medulla), functional units (glomeruli, tubules), and cells (podocytes, mesangial cells in glomerulus). Prior studies have predominantly overlooked the intricate spatial interrelations among objects from clinical knowledge. In this research, we introduce a novel universal proposition learning approach, called panoramic renal pathology segmentation (PrPSeg), designed to segment comprehensively panoramic structures within kidney by integrating extensive knowledge of kidney anatomy. In this paper, we propose (1) the design of a comprehensive universal proposition matrix for renal pathology, facilitating the incorporation of classification and spatial relationships into the segmentation process; (2) a token-based dynamic head single network architecture, with the improvement of the partial label image segmentation and capability for future data enlargement; and (3) an anatomy loss function, quantifying the inter-object relationships across the kidney.

Normalization and batch correction are critical steps in processing single-cell RNA sequencing (scRNA-seq) data, which remove technical effects and systematic biases to unmask biological signals of interest. Although a number of computational methods have been developed, there is no guidance for choosing appropriate procedures in different scenarios. In this study, we assessed the performance of 28 scRNA-seq noise reduction procedures in 55 scenarios using simulated and real datasets. The scenarios accounted for multiple biological and technical factors that greatly affect the denoising performance, including relative magnitude of batch effects, the extent of cell population imbalance, the complexity of cell group structures, the proportion and the similarity of nonoverlapping cell populations, dropout rates and variable library sizes. We used multiple quantitative metrics and visualization of low-dimensional cell embeddings to evaluate the performance on batch mixing while preserving the original cell group and gene structures. Based on our results, we specified technical or biological factors affecting the performance of each method and recommended proper methods in different scenarios. In addition, we highlighted one challenging scenario where most methods failed and resulted in overcorrection. Our studies not only provided a comprehensive guideline for selecting suitable noise reduction procedures but also pointed out unsolved issues in the field, especially the urgent need of developing metrics for assessing batch correction on imperceptible cell-type mixing.

Very short tandem repeats bear substantial genetic, evolutional, and pathological significance in genome analyses. Here, we compiled a census of tandem mono-nucleotide/di-nucleotide/tri-nucleotide repeats (MNRs/DNRs/TNRs) in GRCh38, which we term "polytracts" in general. Of the human genome, 144.4 million nucleotides (4.7%) are occupied by polytracts, and 0.47 million single nucleotides are identified as polytract hinges, i.e., break-points of tandem polytracts. Preliminary exploration of the census suggested polytract hinge sites and boundaries of AAC polytracts may bear a higher mapping error rate than other polytract regions. Further, we revealed landscapes of polytract enrichment with respect to nearly a hundred genomic features. We found MNRs, DNRs, and TNRs displayed noticeable difference in terms of locational enrichment for miscellaneous genomic features, especially RNA editing events. Non-canonical and C-to-U RNA-editing events are enriched inside and/or adjacent to MNRs, while all categories of RNA-editing events are under-represented in DNRs. A-to-I RNA-editing events are generally under-represented in polytracts. The selective enrichment of non-canonical RNA-editing events within MNR adjacency provides a negative evidence against their authenticity. To enable similar locational enrichment analyses in relation to polytracts, we developed a software Polytrap which can handle 11 reference genomes. Additionally, we compiled polytracts of four model organisms into a Track Hub which can be integrated into USCS Genome Browser as an official track for convenient visualization of polytracts.

Illumina genotyping arrays have powered thousands of large-scale genome-wide association studies over the past decade. Yet, because of the tremendous volume and complicated genetic assumptions of Illumina genotyping data, processing and quality control (QC) of these data remain a challenge. Thorough QC ensures the accurate identification of single-nucleotide polymorphisms and is required for the correct interpretation of genetic association results. By processing genotyping data on > 100 000 subjects from >10 major Illumina genotyping arrays, we have accumulated extensive experience in handling some of the most peculiar scenarios related to the processing and QC of Illumina genotyping data. Here, we describe strategies for processing Illumina genotyping data from the raw data to an analysis ready format, and we elaborate on the necessary QC procedures required at each processing step. High-quality Illumina genotyping data sets can be obtained by following our detailed QC strategies.

BACKGROUND: One of the most important and often neglected components of a successful RNA sequencing (RNA-Seq) experiment is sample size estimation. A few negative binomial model-based methods have been developed to estimate sample size based on the parameters of a single gene. However, thousands of genes are quantified and tested for differential expression simultaneously in RNA-Seq experiments. Thus, additional issues should be carefully addressed, including the false discovery rate for multiple statistic tests, widely distributed read counts and dispersions for different genes. RESULTS: To solve these issues, we developed a sample size and power estimation method named RnaSeqSampleSize, based on the distributions of gene average read counts and dispersions estimated from real RNA-seq data. Datasets from previous, similar experiments such as the Cancer Genome Atlas (TCGA) can be used as a point of reference. Read counts and their dispersions were estimated from the reference's distribution; using that information, we estimated and summarized the power and sample size. RnaSeqSampleSize is implemented in R language and can be installed from Bioconductor website. A user friendly web graphic interface is provided at http://cqs.mc.vanderbilt.edu/shiny/RnaSeqSampleSize/ . CONCLUSIONS: RnaSeqSampleSize provides a convenient and powerful way for power and sample size estimation for an RNAseq experiment. It is also equipped with several unique features, including estimation for interested genes or pathway, power curve visualization, and parameter optimization.

BACKGROUND: Exome sequencing allows researchers to study the human genome in unprecedented detail. Among the many types of variants detectable through exome sequencing, one of the most over looked types of mutation is internal deletion of exons. Internal exon deletions are the absence of consecutive exons in a gene. Such deletions have potentially significant biological meaning, and they are often too short to be considered copy number variation. Therefore, to the need for efficient detection of such deletions using exome sequencing data exists. RESULTS: We present ExonDel, a tool specially designed to detect homozygous exon deletions efficiently. We tested ExonDel on exome sequencing data generated from 16 breast cancer cell lines and identified both novel and known IEDs. Subsequently, we verified our findings using RNAseq and PCR technologies. Further comparisons with multiple sequencing-based CNV tools showed that ExonDel is capable of detecting unique IEDs not found by other CNV tools. CONCLUSIONS: ExonDel is an efficient way to screen for novel and known IEDs using exome sequencing data. ExonDel and its source code can be downloaded freely at https://github.com/slzhao/ExonDel.

BackgroundHeat map and clustering are used frequently in expression analysis studies for data visualization.Simple clustering and heat map can be produced from the "heatmap" function in R language.However, it has some limitations in producing advanced graphics and is not highly customizable.Thus, we developed an R package heatmap 3 which significantly improves the original heatmap by adding more powerful and convenient features and providing a highly customizable interface.