Human Ranking

Use image-ranker (https://github.com/QuentinWach/image-ranker) to rank the matrices, then import the results into the same notebook. To rank the matrices, the predicted Hi-C matrices must be plotted, we used pyGenomeTracks for this, but any other tool is fine too.

Hi-C matrices similarity scores

Run 3DChromatin Replicate (GenomeDISCO, HiCSpectrum, QuASAR-Rep), HiCRep (Python version), ENT3C, Hi-cGAN (Pearson AUC) and HiCExplorer hicFindTADs for the three TAD scores: TAD fraction, TAD fraction exact match and TAD score MSE

Regression Models

Create regression models by running “regressionModels.ipynb.”

Dependencies

The main dependency is the Hi-cGAN repository (https://github.com/joachimwolff/Hi-cGAN) repo, with additional requirements detailed in the notebooks.

Additional Dependencies

• numpy
• pandas
• scikit-learn
• matplotlib
• joblib
• tabpfn
• xgboost
• catboost
• pyGenomeTracks
• hicrep
• hicexplorer
• cooler
• scipy