rctd-py#
GPU-accelerated Robust Cell Type Decomposition (RCTD) for spatial transcriptomics.
rctd-py is a Python reimplementation of the spacexr RCTD algorithm of Cable et al. (2022), with GPU acceleration through PyTorch. It deconvolves spatial transcriptomics pixels (Visium, Visium HD, Xenium, CosMx, MERFISH, Slide-seq, …) into cell type proportions using a single-cell RNA-seq reference, in minutes instead of hours.
Installation
pip install rctd-py, GPU setup and offline (HPC) installs.
Quick start
Run RCTD from Python in three lines, and pick a deconvolution mode.
Tutorials
End-to-end walkthroughs on public data.
Coming from R spacexr
Parameter mapping, agreement with R and how to reproduce an R run.