search

calico / basenji

Sequential regulatory activity predictions with deep convolutional neural networks.
Apache License 2.0
410 stars 126 forks source link

Making Akita predictions for different bin size #183

Closed shu-z closed 1 year ago

shu-z commented 1 year ago

Hi, I'm interested in making Akita predictions for contact maps with exactly 5000bp length bins (instead of 2048bp). How do you suggest going about this? Thank you!

gfudenberg commented 1 year ago

Hi Shu, I think you'd need try to re-specify a model architecture that pools to the desired bin size, then re-do data preprocessing and model training in that setting. Hope that helps!

shu-z commented 1 year ago

Sounds good. Thank you!