adaptyvbio / proteinflow Goto Github PK

Hi @elkoz,

I noticed that when I set entry_type='biounit' and classes_to_exclude=["homomers", "heteromers"] during dataset initialization, a subset of the entries throw a KeyError. This happens for around 1/6th of all proteins in the test set (949/6063).
It doesn't emerge at all when entry_type='chain'.

Hi. Have you consider foldseek for structure cluster?

Hello,

I've noticed that a large portion of your data/utils.py code is borrowed from our repository, sidechainnet (repo, file). I think it was attributed in the past, but you no longer provide attribution or a citation. Would you mind providing attribution, in this case, by linking to our repository (https://github.com/jonathanking/sidechainnet) or paper (https://doi.org/10.1002/prot.26169) and providing my name?

Thank you for your consideration. I'm very happy that I could contribute to your work since ProteinFlow seems quite exciting.

Best,
Jonathan

Hello,

I had created a Google Colab notebook for data analysis, and it was working brilliantly. Thank you for creating such a great tool!

However, my Colab does not work anymore despite zero code changes. pip install proteinflow no longer seems to work. I am attaching the two command lines that can be used to reproduce the error:

Hi @elkoz,

I noticed while visualizing the proteins that setting the classes_to_exclude method is not properly filtering the required classes. It runs with no issues, but it still returns proteins that are homomeric or heteromeric.
It might be removing some of the homomers/heteromers, but I haven't tested if it does.

Here is the code that I used for filtering and a few of the resulting proteins:

ProteinFlow has become one of my favorite bioinformatics tools, thank you for creating it!

I recently attempted to generate a dataset to my specifications but got this error:

Currently, while installing on a M1 Mac, the package is not found since the build is not currently available (osx-arm64)

PackagesNotFoundError: The following packages are not available from current channels:

  - proteinflow

Current channels:

  - https://conda.anaconda.org/conda-forge/osx-arm64
  - https://conda.anaconda.org/conda-forge/noarch
  - https://conda.anaconda.org/bioconda/osx-arm64
  - https://conda.anaconda.org/bioconda/noarch
  - https://conda.anaconda.org/adaptyvbio/osx-arm64
  - https://conda.anaconda.org/adaptyvbio/noarch
  - https://repo.anaconda.com/pkgs/main/osx-arm64
  - https://repo.anaconda.com/pkgs/main/noarch
  - https://repo.anaconda.com/pkgs/r/osx-arm64
  - https://repo.anaconda.com/pkgs/r/noarch

Hi Liza,

I noticed when trying to create a W&B table visualization for the entire dataset that converting the pickle files into pdbs reveals multiple bugs.

Firstly, I got a "UnpicklingError: unpickling stack underflow" from the line "protein_entry = ProteinEntry.from_pickle(pickle_path)" It did not happen with every protein, so when I handled that exception I realized that PDBParser could not properly parse a few of the generate pdb files, throwing out an error in the line "structure = parser.get_structure(pdb_id, target_path)"

The requirement biotite=0.35 may be too out-of-date to fail to support python=3.12

Solving environment: - warning  libmamba Added empty dependency for problem type SOLVER_RULE_UPDATE
failed

LibMambaUnsatisfiableError: Encountered problems while solving:
  - package proteinflow-1.3.6-0 requires biotite 0.35.0, but none of the providers can be installed

Could not solve for environment specs
The following packages are incompatible
├─ pin-1 is installable and it requires
│  └─ python 3.12.* , which can be installed;
└─ proteinflow is not installable because it requires
   └─ biotite 0.35.0  but there are no viable options
      ├─ biotite 0.35.0 would require
      │  └─ python >=3.10,<3.11.0a0 *_cpython, which conflicts with any installable versions previously reported;
      ├─ biotite 0.35.0 would require
      │  └─ python >=3.11,<3.12.0a0 *_cpython, which conflicts with any installable versions previously reported;
      ├─ biotite 0.35.0 would require
      │  └─ python >=3.8,<3.9.0a0 *_cpython, which conflicts with any installable versions previously reported;
      └─ biotite 0.35.0 would require
         └─ python >=3.9,<3.10.0a0 *_cpython, which conflicts with any installable versions previously reported.

Really nice package! One thing I feel is missing is being able to split based on a set of sequences, for example, sequences that may have some biophysical properties one is trying to predict using ML methods.

I did not find a way to do this if it already exists.

ProteinDataset construction works at the first try, but it throws an error if I try to reconstruct it. I noticed this in Colab, but it would create a bug in a deep learning pipeline where multiple training runs can be initiated in the same machine.

Dear authors,

thanks a lot for this great pipeline! I was able to run some tests without many problems! I have a question though, is it possible to create the actual PDBs behind the pickle files within ProteinFlow?

Maybe I missed it but the filtering seems to indicate nicely how many structures were removed for what reason but not as obviously how many we had initially / in the end?

Great appreciation for your work! Thank you!

Some description

Hey there! Does the antibody dataset include nanobodies? I have downloaded and processed the data but didn't find nanobodies... Maybe we need an additional key to denote whether it is an nanobody.