Test whether we can accurately predict the gene expressions for drugs that haven't been observed during training.
This has nothing to do with pretraining, but answers the question: Do our molecular embeddings allow us to accurately predict gene responses for unseen drugs?
Test whether we can accurately predict the gene expressions for drugs that haven't been observed during training. This has nothing to do with pretraining, but answers the question: Do our molecular embeddings allow us to accurately predict gene responses for unseen drugs?