HERGAST

HERGAST employs a Divide-Iterate-Conquer framework to identify spatial structures and amplify gene expression signals in ultra-large-scale and ultra-high-resolution spatial transcriptomics data.

Key Features:

• Divide-Iterate-Conquer Framework: HERGAST implements a divide-and-conquer strategy specifically devised for spatial transcriptomics data analysis that can extend other computational methods to ultra-large-scale ST data analysis.
• Heterogeneous Graph Network: HERGAST constructs a heterogeneous graph network to incorporate both local and global spatial relationships, addressing potential over-smoothing problems arising from data splitting.
• Spatial Clustering: HERGAST performs high-precision spatial clustering that identifies specific cell populations such as SPP1+ macrophages intermingled within colorectal tumors.
• Gene Expression Signal Amplification: HERGAST enhances gene expression signals to reveal unique spatial expression patterns of key genes in tissues such as breast cancer.

Scientific Applications:

• Colorectal Tumor Analysis: HERGAST identifies SPP1+ macrophages intermingled within colorectal tumors through high-precision spatial clustering.
• Breast Cancer Research: HERGAST reveals unique spatial expression patterns of key genes in breast cancer through enhanced gene expression signals.
• Ultra-Large-Scale Spatial Transcriptomics: HERGAST enables spatial structure identification and signal amplification in ultra-large-scale and ultra-high-resolution spatial transcriptomics datasets that exceed the scale manageable by existing methods.

Methodology:

HERGAST uses a Divide-Iterate-Conquer framework with a heterogeneous graph network that incorporates both local and global spatial relationships to address over-smoothing from data splitting. Performance was evaluated using adjusted rand index (ARI) in simulations and applied to real-world datasets including colorectal tumors and breast cancer.