SHARE-Topic is a Bayesian model used to infer latent representations of cells, chromatin regions, and genes. SHARE-Topic is implemented using a Gibbs sampler.
The code is written specifically to be run on GPU to decrease the running time of an MCMC chain for a given number of topics.
An example on of the analysis carried from the SHARE-Topic output is done on B-lymphoma 10x Multiome dataset.
Download the processed data from https://zenodo.org/records/10418760 to the "data" folder.
To run SHARE_topic in development version, use pip install -e . to load SHARE_topic as a module