Genetic Engineering Attribution Challenge

This repository contains the code for the Genetic Engineering Attribution Challenge. The objective of this challenge is to predict the laboratory of origin for plasmid DNA sequences.

The competition saw 1211 competitors and our proposed CNN model was ranked 14 in the private leaderboard for the competition with a score of 0.9128 on the testset.

Note: Please look at the competition website for the data format

Executing the program

1. Download data for the competition

2. Configure the data directory and other desired parametes in the utils/config.py file

3. Create n folds of the training data for K-fold Cross-Validation

   python utils/create_fold.py
   

4. Train the model

   python engine.py
   

5. Evaluate the model on validation set for n-folds

   python pred_val.py
   

6. Make predictions for test_set

   python pred_test.py
   

7. Create submission file

   python create_submission.py
   

   Note: The submission script finds the top10 labels and assigns equal probability among them to reduce the final file size.

Contact

For more details, please contact:

• Seyedmostafa Sheikhalishahi: [email protected]
  
• Vevake Balaraman: [email protected]