Cardiogenesis Repo This repo contains scripts for the single cell analysis and BPNet model training and analysis for the Cardiogenesis dataset

NOTE: We recommend using the latest version of ChromBPNet https://github.com/kundajelab/chrombpnet to train similar neural network models on custom chromatin accessibility datasets. The model training and analysis code in this repo is deprecated and not supported any longer. It is provided here to reproduce the results of this specific paper.

1. The single cell analysis code is based off ArchR (https://www.archrproject.com/index.html) and the notebooks contain modifications to standard archr functions when they are used in the manuscript.
2. The BPNet code base uses two other repos : KerasAC (https://zenodo.org/record/4248179#.X8skj5NKiF0) and seqdataloader (https://zenodo.org/record/3771365#.X8skqZNKiF0) as part of the data processing and model training scripts. We recommend using the latest version of ChromBPNet https://github.com/kundajelab/chrombpnet to train similar neural network models on custom chromatin accessibility datasets. The model training and analysis code in this repo is deprecated and not supported any longer. It is provided here to reproduce the results of this specific paper.
3. BPNet models are available at Zenodo at https://doi.org/10.5281/zenodo.6789181
4. The cleaned scRNA object is available in zenodo - human6812and19mergedfinalcleaned.rds https://zenodo.org/record/7392252
5. The Fetal heart invivo ArchR object is available at Zenodo - https://zenodo.org/record/7609326
6. The Invitro differentiation ArchR object is available at Zenodo - https://zenodo.org/record/7613031
7. The folders are arranged for different analysis presented in the manuscript a) BPNet - contains all scripts and files for training the BPNet models and scoring the congenital heart disorder mutations b) MotifFiltering - contains all scripts and files required for cleaning the naive motif calls using BPNet models c) OptimalTransport - contains all scripts and files required for trajectory analysis from the fetal atlas d) Rnotebooks - contains scripts and code for all ArchR based single cell analysis and the in vivo to in vitro nearest neighbour analysis