ADEPT: Autoencoder with differentially expressed genes and imputation for robust spatial transcriptomics clustering.

Area of Science:

• Spatial transcriptomics
• Bioinformatics
• Computational biology

Background:

• Spatial transcriptomics (ST) enables gene expression analysis within complex tissues.
• Existing clustering methods struggle with data quality variations across ST platforms.
• Robust analysis requires integrating spatial context and transcriptional profiles.

Purpose of the Study:

• To develop a robust clustering framework for spatial transcriptomics data.
• To address challenges posed by data quality and platform variability in ST analysis.
• To improve the accuracy and reliability of spatial domain identification and other ST analyses.

Main Methods:

• Developed ADEPT, a graph-based, multi-stage clustering framework.
• Utilized a graph autoencoder backbone for data quality control and stabilization.
• Employed iterative clustering on imputed, differentially expressed gene matrices to minimize variance.

Main Results:

• ADEPT demonstrated superior performance compared to existing methods.
• Outperformed other methods across various ST datasets and platforms.
• Showcased effectiveness in spatial domain identification, visualization, trajectory inference, and data denoising.

Conclusions:

• ADEPT provides a robust and versatile solution for spatial transcriptomics data clustering.
• The framework effectively harnesses spatial context and transcriptional information.
• ADEPT enhances the reliability of downstream analyses in spatial transcriptomics.