stSCI: A multi-task learning framework for integrative analysis of single-cell and spatial transcriptomics data

Public summary

* This study presents a computational method stSCI for integrating single-cell transcriptomics with spatial transcriptomics.

* Experimental results demonstrate that stSCI outperforms 27 state-of-the-art methods across multiple datasets for the major transcriptomics data analysis tasks.

* stSCI is capable of analyzing both imaging- and sequencing-based spatial transcriptomics across diverse tissues, organs, and species.

* stSCI resolves the dynamic spatiotemporal response of a lymphatic niche during Salmonella infection.

Abstract

Spatial transcriptomics (ST) preserves spatial context in gene expression analysis yet faces limitations like low resolution and RNA capture inefficiency. To address these, we present stSCI, a computational method integrating single-cell (SC) and ST data into a unified, batch-corrected embedding space. stSCI employs a fusion module with three specialized optimization tasks to generate biologically preserved joint latent representations, enabling five key analyses: spatial domain identification in single/multi-slice ST data, ST deconvolution predicting cell type proportions in low-resolution spots, SC spatial coordinate reconstruction using ST references, and crossmodality batch correction. Evaluated on 13 different ST datasets spanning sequencing- and imaging-based platforms, and benchmarked against 27 state-of-the-art methods, stSCI improves spatial domain identification, maps cell type proportions in ST data, accurately reconstructs tissue architecture and regional structures, and integrates SC/ST datasets by removing batch effects without compromising biological signals. In a key application, stSCI successfully resolves the dynamic spatiotemporal response of a lymphatic niche during Salmonella infection, demonstrating its power to generate novel biological insights from complex disease models. stSCI’s robustness and versatility make it a powerful tool for uncovering tissue organization and molecular functions.

Introduction

The advent of spatial transcriptomics (ST) technologies has transformed our understanding of tissue architecture. These emerging technologies enable researchers to map gene expression profiles directly to specific tissue locations, providing critical insights that traditional methods such as bulk RNA sequencing (RNA-seq) and single-cell RNA-seq (scRNA-seq) cannot achieve.

Current ST platforms are broadly categorized into sequencing-based and imaging-based methods, each with distinct limitations that affect computational analyses. Sequencing-based platforms, such as 10X Visium, offer comprehensive transcriptome coverage by capturing gene expression data from thousands of predefined spots per tissue section. However, their resolution is limited, as each spot often contains transcripts from multiple cells, reducing the ability to resolve cell-specific transcriptional patterns. This trade-off impacts the quality and resolution of the data, influencing the effectiveness of downstream analyses. Conversely, imaging-based methods like MERFISH, seqFISH, and 10X Xenium achieve subcellular resolution but are restricted to predefined gene panels, typically profiling hundreds to over a thousand genes. This limitation in gene coverage makes it challenging for imaging-based approaches to provide the full transcriptomic insights available with sequencing-based platforms. These constraints, including resolution, spot size, and transcriptome coverage, inevitably hinder the optimal performance of computational tools designed for ST data.

As the ST data often suffer from challenges such as limited resolution and incomplete transcriptome coverage, single-cell (SC) transcriptomics, a more mature sequencing technology, provides comprehensive transcriptomic data with SC resolution. Integrating these two complementary data types through advanced computational methods can significantly enhance the quality and scope of downstream analyses. SC data, with its detailed and high-resolution transcriptional profiles, can address the limitations of ST data by compensating for its lower resolution and incomplete coverage. Meanwhile, ST data can contribute valuable spatial context to SC data. This synergistic integration enables a more holistic understanding of tissue architecture and molecular functions, bridging the strengths of both approaches to provide a SC-level perspective on complex biological systems.