/rnakinet-pub

Primary LanguagePythonOtherNOASSERTION

RNAkinet - publication

RNAkinet is a project dedicated to detecting 5EU-modified reads from the raw nanopore sequencing signal. This repository contains the code for tha analysis of the RNAkinet publication. For the actual RNAkinet code refer to https://github.com/maragkakislab/rnakinet

Custom training

  1. Activate a conda environment with snakemake installed (you can use the snakemake.yaml file to create it)
  2. Navigate to the rnakinet/workflow folder
  3. Change the paths and parameters in config/training_setup_custom.py to reflect your data and requirements (provide genome fasta file, paths to fast5s etc...)
  4. Open the Snakemake file and make sure the experiment name is the same one you specified in the config/training_setup_custom.py file
  5. While in the workflow folder, run snakemake --cores 32 --use-conda -np to get a plan for execution. Once ready remove the -np flag to run training
  6. Once training is finished, the model checkpoint will be available in the checkpoints_pl folder
  7. You can use the checkpoint to run the scripts/inference.py to use it for prediction on other fast5 files