データベースファイルを S3 に保存して Mountpoint for Amazon S3 を利用した OpenFold の実行は実用的か検証してみた

# 環境変数の設定
IMAGE_PATH="/mnt/s3/images"
DATABASE_PATH2="/mnt/s3/data"
CURRENT_DIR=$PWD

cd /home/ubuntu/openfold/examples/monomer
mkdir results-s3

# Apptainer 実行コマンド
nohup apptainer exec --nv \
    --bind ${CURRENT_DIR}:/local,${DATABASE_PATH2}:/database \
    ${IMAGE_PATH}/openfold.sif \
    python3 /opt/openfold/run_pretrained_openfold.py \
    /local/fasta_dir \
    /database/pdb_mmcif/mmcif_files/ \
    --uniref90_database_path /database/uniref90/uniref90.fasta \
    [コマンド省略] \
    --output_dir /local/results-s3 > s3.log &`

[2025-01-03 08:48:37,146] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
INFO:/opt/openfold/openfold/utils/script_utils.py:Loaded OpenFold parameters at /database/openfold_params/finetuning_ptm_2.pt...
INFO:/opt/openfold/run_pretrained_openfold.py:Generating alignments for 6KWC_1...
ERROR:root:HHblits failed. HHblits stderr begin:
ERROR:root:- 09:20:09.396 INFO: Searching 65983866 column state sequences.
ERROR:root:HHblits stderr end
Traceback (most recent call last):
  File "/opt/openfold/run_pretrained_openfold.py", line 493, in <module>
    main(args)
  File "/opt/openfold/run_pretrained_openfold.py", line 291, in main
    precompute_alignments(tags, seqs, alignment_dir, args)
  File "/opt/openfold/run_pretrained_openfold.py", line 111, in precompute_alignments
    alignment_runner.run(
  File "/opt/openfold/openfold/data/data_pipeline.py", line 534, in run
    hhblits_bfd_unirefclust_result = run_msa_tool(
  File "/opt/openfold/openfold/data/data_pipeline.py", line 262, in run_msa_tool
    result = msa_runner.query(fasta_path)[0]
  File "/opt/openfold/openfold/data/tools/hhblits.py", line 160, in query
    raise RuntimeError(
RuntimeError: HHblits failed
stdout:


stderr:
- 09:20:09.396 INFO: Searching 65983866 column state sequences.

top - 22:06:23 up  2:34,  3 users,  load average: 4.07, 4.00, 3.38
Tasks: 311 total,   1 running, 310 sleeping,   0 stopped,   0 zombie
%Cpu(s):  0.2 us,  0.2 sy,  0.0 ni, 87.1 id, 12.4 wa,  0.0 hi,  0.0 si,  0.0 st
MiB Mem : 26.1/61912.1  [||||||||||||||||||||                                                         ]
MiB Swap:  0.0/0.0      [                                                                             ]

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
   1392 ubuntu    20   0 5234596   1.5g   9600 S   4.7   2.5  12:07.91 /usr/bin/mount-s3 --allow-delet+
   2322 ubuntu    20   0 1866.3g  16.3g   4.8g D   1.0  27.0   0:21.50 /opt/conda/bin/hhblits -i /loca+
    632 root      20   0  548336  23168   8704 S   0.3   0.0   0:34.06 /usr/bin/nv-hostengine -n --ser+
    651 root      20   0 2317660  44008  31488 S   0.3   0.1   0:05.09 /usr/bin/containerd

[2025-01-04 03:32:58,129] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
INFO:/opt/openfold/openfold/utils/script_utils.py:Loaded OpenFold parameters at /database/openfold_params/finetuning_ptm_2.pt...
INFO:/opt/openfold/run_pretrained_openfold.py:Generating alignments for 6KWC_1...