The Compositional Perturbation Autoencoder (CPA) is a deep generative framework to learn effects of perturbations at the single-cell level. CPA performs OOD predictions of unseen combinations of drugs, learns interpretable embeddings, estimates dose-response curves, and provides uncertainty estimates.
I obtained consistent embedding space visualizations but pretty different numbers on the evaluation metrics. Could you please help advise if I missed anything? Thanks!
Hello,
Thank you for sharing the tutorial notebook! I was trying to replicate the results in the Norman notebook: https://github.com/theislab/cpa/blob/main/docs/tutorials/Norman.ipynb.
I obtained consistent embedding space visualizations but pretty different numbers on the evaluation metrics. Could you please help advise if I missed anything? Thanks!
Results in the tutorial:
![image](https://github.com/theislab/cpa/assets/33033242/7179c1e8-2a10-432f-92f0-3f5511bb0d55)
My results:
![image](https://github.com/theislab/cpa/assets/33033242/8d220baa-36b8-44dd-ba71-3286b6dfb5bc)