Check out the python package and documentation:
github.com/rajewsky-lab...
rajewsky-lab.github.io/pytrance/
Using a GNN trained on transcript neighborhood graphs, pyTrance computes embeddings that encode spatial RNA localization. RNAs with similar embeddings are likely to co-localize.
I'm very excited to announce that my PhD work is out on biorxiv!
pyTrance is a computational method to predict and quantify subcellular RNA co-localization from spatial transcriptomics data.
www.biorxiv.org/content/10.6...
Big thanks to @cledicj.bsky.social , @nukappa.bsky.social , Nikolaus Rajewsky and the whole Rajewsky lab for the support throughout the project!
Whats most exciting: In a Xenium mouse brain dataset we predicted four GABAergic markers (Gad1, Gad2, Hapln1, Kcnmb2) to co-localize in neuronal projections. Experimental validation confirmed our predictions.
(Col19a1 is used as negative control bellow)