CREsted (Cis-Regulatory Element Sequence Training, Explanation, and Design) is an easy-to-use deep learning package for training enhancer models on single-cell ATAC sequencing (scATAC-seq) data. CREsted provides comprehensive analyses and tutorials to study enhancer codes and the ability to design synthetic enhancer sequences at a cell type-specific, nucleotide-level resolution. Integrated into the scverse framework, CREsted is compatible with outcomes from established scATAC-seq processing tools. It employs novel scATAC-seq preprocessing techniques, such as peak height normalization across cell types, offers flexibility and variety in deep learning modeling architectures and tasks, and contains thorough analysis of the cell type-specific enhancer codes captured during modeling that can also be used for the design of synthetic sequences.
Please refer to the documentation. In particular, the
You need to have Python 3.9 or newer installed on your system and a deep learning backend to be able to use CREsted.
CREsted is build on top of keras 3 and can therefore be used with your deep learning backend of choice (Tensorflow or Pytorch).
pip install tensorflow[and-cuda]
# or
pip install torch
crested
from PyPIpip install crested
pip install crested[tfmodisco]
This requires a cmake installation on your system. If you don't have it, you can install it with:
pip install cmake
See the changelog.
For questions and help requests, please use the issue tracker.
Kempynck, N., Mahieu, L., Ekşi, E. C., Konstantakos, V., Blaauw, C., De Winter, S., Hulselmans, G., Taskiran, I., & Aerts, S. (2024). CREsted: Cis Regulatory Element Sequence Training, Explanation, and Design (1.1.0). Zenodo. https://10.5281/zenodo.13918932
CREsted is build on top of keras 3.0 and can therefore be used with your deep learning backend of choice (Tensorflow or Pytorch). If you don't have a preference, you can take the following into account: