DeepST

DeepST identifies spatial domains in spatial transcriptomics (ST) datasets using deep learning to delineate regions with similar gene expression and histological characteristics.

Key Features:

• Deep Learning Framework: Employs a deep learning architecture tailored for spatial domain identification in ST data.
• Benchmark Performance: Demonstrated superior performance on benchmarking datasets of the human dorsolateral prefrontal cortex, outperforming current methodologies.
• Application in Cancer Research: Applied to breast cancer ST datasets to enable finer-scale dissection of spatial domains within cancer tissues.
• Batch Integration: Integrates ST data from multiple batches or different technologies to harmonize analyses across datasets.
• Expandability for Other Spatial Omics Data: Extensible to process various types of spatial omics data beyond transcriptomics.

Scientific Applications:

• Tissue organization and function: Maps spatial gene expression to resolve tissue organization and function.
• Cancer tissue analysis: Investigates tissue architecture in diseases such as cancer using spatial domain delineation.
• Developmental biology: Studies spatial gene expression patterns during development.
• Pathology: Characterizes spatial features relevant to pathological assessment.
• Regenerative medicine: Informs regenerative medicine studies by identifying spatial domains relevant to tissue repair and organization.

Methodology:

Uses deep learning algorithms to analyze spatial transcriptomics data by identifying regions with similar gene expression profiles and histological characteristics and integrates data across batches and technologies.