Genomic deep learning models used in tangermeme paper.

Five genomic deep learning models used as part of "tangermeme: A toolkit for understanding cis-regulatory logic using deep learning models." Each model makes predictions for a different readout in K562. The ChromBPNet models and ProCapNet models are from the ENCODE Compendium and the DeepSEA/Beluga model is from Kipoi. A model definition file is included for Beluga that resolves the bespoke alphabet assumed by the authors and corrects the AGCT ordering to the more standard ACGT one. See the tutorials on the tangermeme page for how to load and use these models to reproduce the figures in the paper.