Randomized Spatial PCA (RASP): A computationally efficient method for dimensionality reduction of high-resolution...

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Spatial transcriptomics (ST) reveals gene expression in tissue context.
• Understanding spatial domains is crucial for tissue architecture and disease research.
• Existing methods can be computationally intensive for large ST datasets.

Purpose of the Study:

• Introduce Randomized Spatial PCA (RASP), a novel dimensionality reduction technique for ST data.
• Enhance computational speed and scalability for analyzing large-scale ST datasets.
• Provide a flexible method for de-noising and spatially smoothing gene expression.

Main Methods:

• RASP utilizes a randomized two-stage PCA framework.
• Configurable spatial smoothing is integrated into the method.
• Performance was benchmarked against existing ST analysis tools using diverse datasets.

Main Results:

• RASP achieves comparable or superior accuracy in tissue-domain detection compared to existing methods.
• RASP offers significant improvements in computational speed and scalability.
• The method enables efficient exploration of spatial-smoothing parameters for optimal results.

Conclusions:

• RASP provides a computationally efficient and accurate approach for spatial transcriptomics data analysis.
• Its speed and scalability make it suitable for large, high-resolution datasets.
• RASP empowers researchers to better investigate complex tissue architecture and spatial gene expression patterns.