Open miquelduranfrigola opened 8 months ago
@miquelduranfrigola
I've made a small refactoring to the 0_preparation folder: users should run the 0_precalculate_hub_descriptors.sh file to get all the descriptors, including the ersilia embeddings. I have not deleted the original piece of code yet just in case you prefer to keep it
Now I've hardcoded three descriptors, we can make it more user friendly if we want.
I am adding requirements in a requirements file that I'll fill in as needed
Hi @miquelduranfrigola and @AnnaMontaner:
I have tested the pipeline to train pairs of models (to save time I have not tested all the combinations, but they should work). I have made minor refactoring in the code @miquelduranfrigola, to make sure the folders are created when needed. I only have two comments:
bimodal_model.py
. It might be best to have an embeddings_pairs and embeddings_triplets to avoid overriding results - haven't modified it though3_run_prediction_experiment.py
expects all the model combinations to be available. Shouldn't we have them as flags as well, if the rest of the scripts allow for selecting only a few models? this way we don't have to train everything to get results. I am not sure if I am missing anything and we need the results of ALL the combinations.@AnnaMontaner if you try it, I'm using Py3.10 and the package requirements along with the versions is in requirements.txt
Thanks @GemmaTuron - I am aware I am delayed with this. It is well noted and will respond ASAP.
I have done major updates on the pipeline. We may want to run it one last time.
Hi @GemmaTuron
As you know, I have refactored the drug-gene pair pipeline. We should test it and:
requirements.txt
file installing all necessary dependencies.If you encounter any problem, please let me know.