Scripts used to perform calculations including two-step dimension reduction and cluster analysis. Please cite: https://doi.org/10.1186/s13059-024-03472-8.
- alabtools (https://github.com/alberlab/alabtools)
- pickle
- numpy
- scipy
- keras
- tensorflow
- sklearn
- matplotlib
- seaborn
Models generated by the Integrated Genome Modeling Platform (IGM) (https://github.com/alberlab/igm) (saved as for example GM_igm-model.hss) and partition files generated by the Markov Clustering Algorithm (MCL) (https://github.com/GuyAllard/markov_clustering) (saved as for example GM_speckle_clusters.dat) should be placed under directory Model/ for further usage. For the GM12878 model and the generation of its partition files, please refer to https://doi.org/10.5281/zenodo.7352276. The genome populations of H1-hESC and HFFc6 are available at https://doi.org/10.5281/zenodo.7563402.
Script used to perform embedding:
python embedding.py <cell type> <chromosome index> <tag> <starting index> <ending index>
where "cell type" can be GM, H1 or HFF (GM for GM12878, H1 for H1-hESC and HFF for HFFc6), "chromosome index" should be ranging from 1 to 23 (1 to 22 for H1 or HFF), "tag" is 0 or 1 depends on whether autoencoder has been performed or not, "starting index" and "ending index" indicate the bead range.
The following command line generates embedding results for the whole chromosome 6 (bead index 0 to 854) of GM12878:
python embedding.py GM 6 0 0 855
Script used to perform clustering analysis:
python analysis.py <cell type> <chromosome index> <starting index> <ending index>
Similarly, we can run an analysis based on results generated by the embedding:
python analysis.py GM 6 0 855
The results will be generated under directory GM_Structural_Features/. The speckle related calculations are only available for GM12878. To run the analysis without speckle related calculations for GM12878, H1-hESC or HFFc6:
python analysis_simple.py GM 6 0 855
The results will be generated under directory GM/.