The environments:

• All the prediction and deepshap with jupyter_tf (in environment.yml), and everything TF-modisco related with meme_env (meme.yml)
  
  

The models:

• I have the zip files of all the models (the name should be with day e.g. D5 and domain e.g. pMN)
• To use the model files, I have a script maybe_predict.py that saves the prediction in a csv file
  
  
• The training example is in Training/example_usage.txt

Deep explainer:

• deep_explainer/run_deepexplain_savenpz_corrected.py saves the contribution scores of the sequences
• Can be used to visualisation
  
  

tf-modisco:

• Requires npz file of the one hot encoded fasta
• Command line modisco motifs in script modisco_and_meme/run_modisco_int_new1.sh, then compared to jaspar database using modisco_and_meme/submit_meme2.sh
• To help produce understandable reports modisco_and_meme/Interpret_modisco.ipynb is used to extract motif names and find matching names from MA codes

Interaction analysis:

• In the folder interaction_analysis
• Bash script to run in interaction_analysis/interactiveD11.txt
• (Implementation requires understanding the orientation of motif, impact of placement of motifs, distances between motifs)