Skip to content

Latest commit

 

History

History
117 lines (69 loc) · 1.99 KB

README.md

File metadata and controls

117 lines (69 loc) · 1.99 KB
Raw

General

save_hf_model_locally.py

  • Huggingface models must be saved locally to run on the cluster
  • argparse: TODO

fasta2spacedtable.py

  • Convert a fasta to table of ID,S E Q U E N C E
  • argparse

hf_util.py

  • Common utilies
  • More to go in here, commonalities between sequence/aa level classification

hf_similarity.py

  • get similarity scores for fasta of proteins
  • argparse
  • logger: TODO

hf_aligner.py

  • get amino acid similarities

Classify whole sequences

hf_classification.py

  • Classify labeled sequences (ex. chloroplast localization)
  • argparse
  • logger
  • evaluator: hf_evaluation.py

hf_evaluation.py

  • Evaluate performance of sequence classification
  • argparse
  • logger: TODO

hf_predict.py

  • Role potentially filled by hf_interpret.
  • Apply classification model to new sequences: TODO
  • argparse: TODO
  • logger: TODO

hf_interpret.py

  • Apply classification and get aa level explainations
  • argparse
  • logger: TODO

hf_embed.py

  • Convert fasta to embedding
  • argparse
  • logger: TODO

hf_cluster.py

  • Cluster sequence embeddings
  • argparse: TODO
  • logger:TODO

hf_interpret.py

  • Get word importances for each sequence classification
  • argparse: TODO
  • logger:TODO

Classify amino acids

hf_classification_aa.py

  • Classify individual amino acids
  • argparse
  • logger
  • evaluator: TODO

hf_evaluate_aa.py : TODO

  • Evaluate performance of sequence classification
  • argparse: TODO
  • logger: TODO

hf_predict_aa.py: TODO

  • Apply classification model to new sequences: TODO
  • argparse: TODO
  • logger: TODO

Figure out what to do for interpretibility

Motifs from pretrained prot-bert

for each amino acid... Get embedding Get saliency map of each sequence https://colab.research.google.com/github/AndreasMadsen/python-textualheatmap/blob/master/notebooks/huggingface_bert_example.ipynb#scrollTo=X8GJbpoUmYdT

for each aa record saliency interactions, and embeddings of those interactions

compare similarity of embeddings.....