phbradley
commented
1 year ago Hi there,
Thanks for trying the code! Yes, there are some compatibility issues with alphafold and jax. Have you tried using an older version of jax, for example, 0.2.22, and for jaxlib 0.1.72 ? I've added a google colab notebook that shows a working example of installing the software in the colab setting:
https://github.com/phbradley/TCRdock/blob/main/tcrdock_colab_pipeline_v1.ipynb
which uses the versions listed here:
https://github.com/phbradley/TCRdock/blob/main/requirements_colab_python38.txt
I also don't think it would be too hard to use a more recent version of alphafold-- as you note below, the changes we made are very minimal, so you could also try that route.
Take care, Phil
From: Su Wu @.> Sent: Wednesday, August 16, 2023 10:44 AM To: phbradley/TCRdock @.> Cc: Subscribed @.***> Subject: [phbradley/TCRdock] Internal alphafold library issue (Issue #6)
CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.
Hello, I followed the suggested command lines on the git to run TCRdock.
Step 1: Parse a TCR:peptide-MHC ternary structure PDB file
python parse_tcr_pmhc_pdbfile.py --pdbfiles examples/parsing/1qsf.pdb --organism human --mhc_class 1 --out_tsvfile output/parse_output.tsv
Step 2: Set up for AlphaFold modeling a set of TCR:pMHC complexes
python setup_for_alphafold.py --targets_tsvfile examples/benchmark/single_target.tsv --output_dir test_setup_single
Step 3: Run AlphaFold modeling
python run_prediction.py --targets test_setup_single/targets.tsv --outfile_prefix test_run_single --model_names model_2_ptm --data_dir /n/shared_db/alphafold/
Unfortunately, I encountered the following error message:
(tcrdock_test_python3.9) @.*** TCRdock]$ python run_prediction.py --targets test_setup_single/targets.tsv \
--outfile_prefix test_run_single --model_names model_2_ptm \ --data_dir /n/shared_db/alphafold/
Traceback (most recent call last): File "/home/sw34/TCRdock/run_prediction.py", line 77, in import predict_utils File "/home/sw34/TCRdock/predict_utils.py", line 14, in import haiku as hk File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/haiku/init.py", line 17, in from haiku import data_structures File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/haiku/data_structures.py", line 17, in from haiku._src.data_structures import to_immutable_dict File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/haiku/_src/data_structures.py", line 30, in from haiku._src import utils File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/haiku/_src/utils.py", line 24, in import jax File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/jax/init.py", line 108, in from .experimental.maps import soft_pmap File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/jax/experimental/maps.py", line 25, in from .. import numpy as jnp File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/jax/numpy/init.py", line 16, in from . import fft File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/jax/numpy/fft.py", line 17, in from jax._src.numpy.fft import ( File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/jax/_src/numpy/fft.py", line 19, in from jax import lax File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/jax/lax/init.py", line 330, in from jax._src.lax.fft import ( File "/home/sw34/.conda/envs/tcrdock_test_python3.9/lib/python3.9/site-packages/jax/_src/lax/fft.py", line 144, in xla.backend_specific_translations['cpu'][fft_p] = pocketfft.pocketfft AttributeError: module 'jaxlib.pocketfft' has no attribute 'pocketfft'
This error seems to be a known issue with Alphafold and was fixed in a newer version of Alphafold on August 9, 2022. deepmind/alphafold#523https://urldefense.com/v3/__https://github.com/deepmind/alphafold/issues/523__;!!GuAItXPztq0!g3UuVRjD7cWrbL4LIGz_LpexrHCrZoYi5jc6W8Ubd2MqmfcbtfJs3SRuXhmU2RnlRNNd2KJHIsTh6ZP4ShnAdy-6$ However, based on the information provided in the changes_to_alphafold.txt file in the TCRdock repository, the snapshot associated with the linked alphafold commit dates back to November 5, 2021.
Would it be feasible to consider updating the internal alphafold library bundled within TCRdock to a more recent version? Thank you very much!
— Reply to this email directly, view it on GitHubhttps://urldefense.com/v3/__https://github.com/phbradley/TCRdock/issues/6__;!!GuAItXPztq0!g3UuVRjD7cWrbL4LIGz_LpexrHCrZoYi5jc6W8Ubd2MqmfcbtfJs3SRuXhmU2RnlRNNd2KJHIsTh6ZP4StJlv6Nc$, or unsubscribehttps://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/ABBNCHZYTRGNUZHPJENFMLLXVUBJBANCNFSM6AAAAAA3S4EFWM__;!!GuAItXPztq0!g3UuVRjD7cWrbL4LIGz_LpexrHCrZoYi5jc6W8Ubd2MqmfcbtfJs3SRuXhmU2RnlRNNd2KJHIsTh6ZP4Sgb0NT7h$. You are receiving this because you are subscribed to this thread.Message ID: @.***>
Hello, I followed the suggested command lines on the git to run TCRdock.
## Step 1: Parse a TCR:peptide-MHC ternary structure PDB file python parse_tcr_pmhc_pdbfile.py --pdbfiles examples/parsing/1qsf.pdb \ --organism human --mhc_class 1 --out_tsvfile output/parse_output.tsv
## Step 2: Set up for AlphaFold modeling a set of TCR:pMHC complexes python setup_for_alphafold.py --targets_tsvfile examples/benchmark/single_target.tsv \ --output_dir test_setup_single
## Step 3: Run AlphaFold modeling python run_prediction.py --targets test_setup_single/targets.tsv \ --outfile_prefix test_run_single --model_names model_2_ptm \ --data_dir /n/shared_db/alphafold/
Unfortunately, I encountered the following error message:
(tcrdock_test_python3.9) [sw34@compute-a-16-162 TCRdock]$ python run_prediction.py --targets test_setup_single/targets.tsv \
This error seems to be a known issue with Alphafold and was fixed in a newer version of Alphafold on August 9, 2022. https://github.com/deepmind/alphafold/issues/523 However, based on the information provided in the
changes_to_alphafold.txtfile in the TCRdock repository, the snapshot associated with the linked alphafold commit dates back to November 5, 2021.Would it be feasible to consider updating the internal alphafold library bundled within TCRdock to a more recent version? Thank you very much!