Add more Embeddings - Githubissues

theislab / chemCPA

Code for "Predicting Cellular Responses to Novel Drug Perturbations at a Single-Cell Resolution", NeurIPS 2022.

https://arxiv.org/abs/2204.13545

MIT License

99 stars 24 forks source link

Add more Embeddings #29

Closed siboehm closed 2 years ago

siboehm commented 2 years ago

[x] 2x pretrain GNN from Dgl LifeSci @MxMstrmn
[X] GROVER_base
[x] rdkit generic feature vector @MxMstrmn
[X] HierVAE / JTVAE. Make sure the encoding issues aren't just due to canonicalization (maybe RDKIT 2019 vs 2021)? @siboehm
[x] Simple Seq2Seq LSTM-based NLP VAE @siboehm
[ ] (optional) Nvidia MegaMolBART

Checklist:

make sure the Index of the dataframe is the SMILES as stored in embeddings/lincs_trapnell.smiles
make sure to add a branch in the embedding function: compert/embedding.py: get_chemical_representation.
make sure the embedding has been moved into the correct location, at /storage/groups/ml01/projects/2021_chemicalCPA_leon.hetzel/embeddings/

siboehm commented 2 years ago

@MxMstrmn: I don't have write access to our global directory /storage/groups/ml01/projects/2021_chemicalCPA_leon.hetzel/embeddings/. I need to copy the seq2seq embeddings as well as the new grover_base.parquet (the df in the global dir turns out to be outdated) into this directory. Both are stored at /mnt/home/icb/simon.boehm/Masters_thesis/chemical_CPA/embeddings.

Can you give me write access or copy them for me? The current large LINCS sweep loads all embeddings from my home dir, due to the outdated grover embedding.

MxMstrmn commented 2 years ago

@siboehm, I wanted to copy them but did not have acces on your embeddings either. I now changed the rights of the project folder /storage/groups/ml01/projects/2021_chemicalCPA_leon.hetzel/embeddings/ to be 775

siboehm commented 2 years ago

Thanks that worked, I updated the embeddings

siboehm commented 2 years ago

Now that we have Junction Tree VAE #49 this can be closed