GPU-accelerated, self-optimizing processing for 3D multiplexed iterative RNA-FISH experiments.

Area of Science:

• Genomics
• Computational Biology
• Microscopy

Background:

• Imaging-based spatial transcriptomics requires high-resolution optics and computational processing.
• Current methods often trade local information for faster, larger-scale imaging by increasing axial sampling distance.

Purpose of the Study:

• Introduce a GPU-accelerated computational framework, merfish3d-analysis, to expedite spatial transcriptomics data processing.
• Quantify information loss associated with axial sampling in imaging-based spatial transcriptomics.
• Improve experimental data quality and accessibility for spatial transcriptomics.

Main Methods:

• Developed and implemented a GPU-accelerated computational framework (merfish3d-analysis).
• Simulated imaging-based spatial transcriptomic experiments to assess information loss from axial sampling.
• Reprocessed publicly available multiplexed error-robust fluorescence in situ hybridization (MERFISH) datasets.
• Performed new MERFISH experiments on a post-mortem human olfactory bulb sample.
• Designed and applied a multi-step autofluorescence quenching protocol.

Main Results:

• The merfish3d-analysis framework significantly speeds up computational processing of spatial transcriptomic data.
• Quantified information loss due to varying axial sampling resolutions in simulated experiments.
• Successfully reprocessed existing MERFISH datasets and analyzed new data from a human olfactory bulb.
• The developed autofluorescence quenching protocol enhanced experimental data quality.

Conclusions:

• The merfish3d-analysis framework accelerates computational processing for imaging-based spatial transcriptomics.
• Understanding axial sampling trade-offs is crucial for optimizing data acquisition.
• Improved sample preparation and accessible computational tools can democratize spatial transcriptomics.