Testing UniversalEPI on HepG2 data

These results are generated using the steps described in the section "Generating prediction on new data".

Performance evaluation

Predicted HepG2 results can downloaded and used to calculate the overall correlation as well as distance-stratified correlation using Stage2/plot_scores.ipynb.

Uncertainty calibration

Using ensembles and estimated aleatoric uncertainty, one can calculate total uncertainty as done in Stage2/plot_scores.ipynb.

Visualizing predictions

Finally, one can visualize the Hi-C predictions using the plot_arcs() function defined in Stage2/plot_scores.ipynb.