GANDALF

This code repo has the code associated with the paper GANDALF: Generative AttentioN based Data Augmentation and predictive modeLing Framework for personalized cancer treatment.

GANDALF is a cancer drug response prediction model that addresses the problem of limited labelled patient data through a novel genomic data augmentation technique.

Folders

• Data contains the notebooks needed to process the raw data files.
• Processing has files for additional processing such as annotations for mutations.
• Model has the relevant files for training GANDALF, along with relevant files for comparison with SOTA, ablation, sensitivity etc.
• Evaluation has notebooks for generating plots and visualizing results.

Running GANDALF

Navigate to Model folder and perform the following steps.

1. Ensure that the correct values are filled up in experiment_settings_yaml/model_config_2A_annotated_mutations_v7_foldX.yaml, especially sample_id, input_data_type, feature_num and wandb_config details.
2. Run step 1 and step 2 of GANDALF using

python model_pretraining.py experiment_settings_yaml/model_config_2A_annotated_mutations_v7_foldX.yaml

3. Use the saved augmented data to run steps 3, 4, 5 by running the notebook mtl_pseudolabeling.ipynb or mtl_pseudolabeling-drug_specific.ipynb as needed.