Results

Computer vision

Test

• ROC AUC 0.697
• MAP 0.183

Graph method

Test

• ROC AUC 0.702
• MAP 0.199

Training models

Computer vision

0. Generate graph images python generate_images.py

1. Prepare data by python prepare_data.py

2. Adjust config in config/config_classification.yml

3. train models run python train.py

4. Watch tensorboad logs tensorboard --logdir ./lightning_logs/

5. Collect up-to-date requirements.txt call pipreqs --force

Graph method

1. Run python fit_predict_graph.py

Data

We will predict the activity (against COVID?) of different molecules. 

Dataset sample:

smiles,activity
OC=1C=CC=CC1CNC2=NC=3C=CC=CC3N2,1
CC(=O)NCCC1=CNC=2C=CC(F)=CC12,1
O=C([C@@H]1[C@H](C2=CSC=C2)CCC1)N,1