Bulk examples and tutorials (L1000) #21
Description
Hi, first of all, thank you for the great work, it looks very exciting!!
I have a question regarding the tutorials. All the examples I found were related to single-cell data. Both on cell type/line prediction, cell generation, and perturbation prediction.
Could you provide examples for prompts you used for the bulk data and more details on L1000?
From the paper and the tutorials, I inferred that the drug perturbation predictions were done by giving the compounds' common names. Have you tried using SMILES or any other general compound representation to be able to generalize to new drugs instead of just known names?
And also it wasn't very clear to me that what data was each model trained on. I am using C2S-Scale-Gemma-27B but I'm not sure if this checkpoint on hugginface has seen L1000 data during pretraining, or it has only seen single-cell and the results are all after a fine-tuning which I should do as well.
It would be great if you could provide more information regarding bulk data.
Thank you,
Arian