ifreecell / ppr-motif

PPR Motif Annotation using Neural Network

PPR Motif Annotation using Neural Network

Overview

This is the code and weight files to run and deploy the Flask app PPR motif, which uses pre-trained Keras model to predict 10 different variants of PPR motif. However, we still use PROSITE model of the PPR motif.

Sequence logo of Arabidopsis thaliana PPR motif variants

P

P1

P2

L1

L2

S1

S2

SS

TPR

E1

E2

Dependencies

sudo pip install -r requirements.txt

Todo

• Bugs
  • Handling inputs.
  • Bed format coordinates (0-based exclusive) of the features.
  • Ending positions of the features.
• Optimization
  • Variable length features (pad_sequences?).
  • Unbalanced training set (sample_weight?).
  • Under-represented classes (class_weight?).
• Enhancement
  • Setting maximum number of query sequences.
  • Loading example query sequences from file.
  • Uploading query sequences from file.
  • Displaying and downloading annotations in either bed or GFF3 format.