Diffusion Correction Benchmarking

[DarioS]

Now that there is public data generated by the new RNA and protein Xenium assay, could the diffusion algorithm be benchmarked? It appears more severe than I imagined that it would be. Blue dots should be in pink circles.

[dcjones]

Yeah this is pretty interesting. Now that protein data is available, some measure of the agreement between RNA and protein before and after trying to correct for diffusion could be quite useful for calibration.

[tjbencomo]

With the paired RNA and protein data, it could be interesting to investigate spurious coexpression of protein markers using Proseg vs default 10x segmentation, like you do in the Proseg paper. Been meaning to do this once I'm a little less busy

[DarioS]

True location of RNAs seems to be a hot topic nowadays.

ELLA is a statistical framework for modeling subcellular mRNA localization and detecting spatially variable genes within cells. ELLA uses an over-dispersed nonhomogeneous Poisson process to model spatial count data with a unified cellular coordinate system to anchor diverse cellular morphologies, demonstrating effective type I error control and high power in simulations. In real data applications, ELLA identifies genes with distinct subcellular localization and associate these patterns to key mRNA characteristics: nuclear-enriched genes exhibit an abundance of long noncoding RNAs or protein-coding mRNAs, while cytoplasmic- or membrane-enriched genes frequently encode ribosomal proteins or contain signal peptides. ELLA also uncovers dynamic subcellular localization changes across the cell cycle. Overall, ELLA is a powerful, robust, and scalable tool for subcellular spatial expression analysis across high-resolution spatial transcriptomics platforms.

Source: ELLA: Modeling Subcellular Spatial Variation of Gene Expression within Cells in High-resolution Spatial Transcriptomics, Nature Communications.

[dcjones]

This looks interesting, thanks for the link!