Trained model checkpoints and evaluation metrics for "Deep learning models for chemical perturbation prediction do not yet utilise drug molecular features"

Trained model weights and per-fold evaluation metrics from a benchmarking comparison of seven published L1000 chemical perturbation models (DeepCE, CIGER, MultiDCP, TranSiGen, PRnet, PertDiT, XPert) and a drug-blind MLP baseline. `dpb_outputs.tar` (≈9.5 GB) contains 364 best-by-validation checkpoints and 364 `all_metrics.json` files, organised by experiment / model / seed / condition.