Use machine learning to design codon-optimized DNA sequences for increased protein expression in CHO cells.

Description of project: https://www.liebertpub.com/doi/10.1089/cmb.2021.0458

Packages required:

• Python (built using Python 3.7.4 via Ubuntu-18.04)
• Keras (tensorflow)
• json
• NumPy
• OS

To predict the 'best' DNA sequence for a given amino acid sequence using the pre-trained model, use rnn_predict.py.

Codon optimization procedure:

1. Place the following in the same directory:

• rnn_predict.py
• rnn_model.h5
• dna_tokenizer.json
• aa_tokenizer.json
• Text file containing amino acid sequence (save as "sequence.txt")
  • Single-letter amino acid abbreviations
  • Signal peptide included, if applicable
  • No stop codon (will be added to output DNA sequence)

2. In line 19 of rnn_predict.py, change to the working directory containing the above files.
3. Run rnn_predict.py.
4. The optimized DNA sequence is output as "sequence_opt.txt" in the same directory.