Peak calling prototypes for DiMeLo-seq.
The purpose of this repository is to test a few approaches for calling binding peaks for various protein binding datasets with Directed Methylation and Long Read Sequencing, DiMeLo-seq.
The scripts in here run the dimelo_arbitrary_basemod_dev branch, https://github.com/OberonDixon/dimelo/tree/arbitrary_basemod_dev, which allows for rapid full-genome pileups and single-read processing. Currently it is set up to run on savio3 nodes on the Berkeley high performance cluster, which allows ~90GB of memory. A smaller bases cap per process would be necessary to run this elsewhere without crashing.
This code is there to give an interface to load bigwigs to arrays from specifically the validMOD and modMOD suffix format used by the dimelo dev package.
The various notebooks use pomegranate, a statistical modeling package, to train HMMs. There is also some simple p-value testing with scipy as a baseline against which to comapre.