Skip to content

Latest commit

 

History

History
11 lines (6 loc) · 652 Bytes

README.md

File metadata and controls

11 lines (6 loc) · 652 Bytes

Basset-style peak prediction in Basenji

Basset: learning the regulatory code of the accessible genome with deep convolutional neural networks. Genome Research 5/2016.

get_dnase.sh - Download DNase peak files from ENCODE and Roadmap.

make_dataset.sh - Recommended parameters to generate a Basset-style single peak annotation dataset.

params_basset.json - Recommended parameters to train a Basset-style model. Note, these aren't the exact parameters used in that paper; they represent a roughly benchmarked improvement.

basenji_train.py -k -o train_basset params_basset.json data_basset