You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Run tests: fail-fast: true to save some GitHub action resources
Restructured loader.py to reduce duplicated code
Gene lists are now in the meta directory of Zenodo to avoid duplication. Had to adapt the loaders to always download the meta/ directory and adapt load_and_select_gene_features to load the gene lists from meta/.
#307 Added pydantic dependency until new ray version because of ray-project/ray#58354: ray.tune automatically imports ray.train which requires pydantic
New features
#283, #293 New DrugGNN Model that uses drug-molecule graphs as inputs (and gene expression for the cell lines)
#284 New ChemBERTaNeuralNetwork that uses ChemBERTa pre-trained embeddings for the drugs
#304 drevalpy can now be installed with an optional ray dependency: pip install drevalpy[multiprocessing]
#305 featurizer scripts for automatically creating drug features for new drugs in drevalpy/datasets/featurizer: ChemBERTa embeddings, drug graphs, and MolGNet embeddings + tests (are right now skipped because transformers and rdkit are not automatically installed)