Space: reconciling multiple spatial domain identification algorithms via consensus clustering

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Spatially resolved transcriptomics (SRT) technologies offer insights into tissue architecture.
• Computational methods are used to identify spatial domains within tissues.
• Inconsistent performance across different algorithms hinders reliable downstream analysis.

Purpose of the Study:

• To develop a robust domain identification method for SRT data.
• To address the challenge of inconsistent results from various computational algorithms.
• To provide a reliable tool for analyzing tissue architecture and biological features.

Main Methods:

• Propose 'Space,' a novel domain identification method for SRT.
• Measure algorithm consistency to select reliable methods.
• Construct a consensus matrix integrating multiple algorithm outputs.
• Incorporate similarity loss, spatial loss, and low-rank loss for accuracy and efficiency.

Main Results:

• Space resolves inconsistent clustering labels from different methods.
• Achieves highly reliable clustering output for spatial domains.
• Demonstrates exceptional performance in deciphering key tissue structures and biological features across multiple SRT datasets.
• Provides flexible interfaces for visualization, gene analysis, and trajectory inference.

Conclusions:

• Space offers a reliable and accurate solution for spatial domain identification in SRT.
• The method enhances the interpretability of tissue architecture and biological insights from SRT data.
• Space is easily installable and available with source code for broader accessibility.