Incorrect Structure Output

[evans-ZH]

Hi, I am running AlphaFold3 using Singularity to predict the structure of 2pv7. However, the output structure appears meaningless. It seems AlphaFold3 may not be running correctly. I wonder how can I solve this problem? Thanks.

I1127 11:32:32.858946 22391989208192 folding_input.py:1044] Detected /root/af_input/fold_input.json is an AlphaFold 3 JSON since the top-level is not a list.
I1127 11:32:36.585006 22391989208192 xla_bridge.py:895] Unable to initialize backend 'rocm': module 'jaxlib.xla_extension' has no attribute 'GpuAllocatorConfig'
I1127 11:32:36.739334 22391989208192 xla_bridge.py:895] Unable to initialize backend 'tpu': INTERNAL: Failed to open libtpu.so: libtpu.so: cannot open shared object file: No such file or directory
2024-11-27 11:32:40.469534: W external/xla/xla/service/gpu/nvptx_compiler.cc:930] The NVIDIA driver's CUDA version is 12.4 which is older than the PTX compiler version 12.6.77. Because the driver is older than the PTX compiler version, XLA is disabling parallel compilation, which may slow down compilation. You should update your NVIDIA driver or use the NVIDIA-provided CUDA forward compatibility packages.
I1127 11:32:44.300118 22391989208192 pipeline.py:40] Getting protein MSAs for sequence GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 11:32:44.305664 22388865492544 jackhmmer.py:78] Query sequence: GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 11:32:44.306400 22388863391296 jackhmmer.py:78] Query sequence: GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 11:32:44.306827 22388624324160 jackhmmer.py:78] Query sequence: GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 11:32:44.307636 22388622222912 jackhmmer.py:78] Query sequence: GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 11:32:44.307905 22388865492544 subprocess_utils.py:68] Launching subprocess "/hmmer/bin/jackhmmer -o /dev/null -A /tmp/tmpctjdkv48/output.sto --noali --F1 0.0005 --F2 5e-05 --F3 5e-07 --cpu 8 -N 1 -E 0.0001 --incE 0.0001 /tmp/tmpctjdkv48/query.fasta /root/public_databases/uniref90_2022_05.fa"
I1127 11:32:44.308365 22388863391296 subprocess_utils.py:68] Launching subprocess "/hmmer/bin/jackhmmer -o /dev/null -A /tmp/tmpqdp6a2xj/output.sto --noali --F1 0.0005 --F2 5e-05 --F3 5e-07 --cpu 8 -N 1 -E 0.0001 --incE 0.0001 /tmp/tmpqdp6a2xj/query.fasta /root/public_databases/mgy_clusters_2022_05.fa"
I1127 11:32:44.309545 22388622222912 subprocess_utils.py:68] Launching subprocess "/hmmer/bin/jackhmmer -o /dev/null -A /tmp/tmpmua0fj8p/output.sto --noali --F1 0.0005 --F2 5e-05 --F3 5e-07 --cpu 8 -N 1 -E 0.0001 --incE 0.0001 /tmp/tmpmua0fj8p/query.fasta /root/public_databases/uniprot_all_2021_04.fa"
I1127 11:32:44.309772 22388624324160 subprocess_utils.py:68] Launching subprocess "/hmmer/bin/jackhmmer -o /dev/null -A /tmp/tmpy9obs77n/output.sto --noali --F1 0.0005 --F2 5e-05 --F3 5e-07 --cpu 8 -N 1 -E 0.0001 --incE 0.0001 /tmp/tmpy9obs77n/query.fasta /root/public_databases/bfd-first_non_consensus_sequences.fasta"
I1127 11:35:18.763229 22388624324160 subprocess_utils.py:97] Finished Jackhmmer in 154.453 seconds
I1127 11:40:31.640248 22388865492544 subprocess_utils.py:97] Finished Jackhmmer in 467.332 seconds
I1127 11:44:29.538374 22388622222912 subprocess_utils.py:97] Finished Jackhmmer in 705.228 seconds
I1127 11:47:57.616615 22388863391296 subprocess_utils.py:97] Finished Jackhmmer in 913.308 seconds
I1127 11:47:57.781585 22388863391296 jackhmmer.py:137] Limiting MSA depth to 5000
I1127 11:47:57.788024 22391989208192 pipeline.py:73] Getting protein MSAs took 913.49 seconds for sequence GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 11:47:57.788283 22391989208192 pipeline.py:79] Deduplicating MSAs and getting protein templates for sequence GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 11:47:57.939319 22388622222912 subprocess_utils.py:68] Launching subprocess "/hmmer/bin/hmmbuild --informat stockholm --hand --amino /tmp/tmpb3h8bf7w/output.hmm /tmp/tmpb3h8bf7w/query.msa"
I1127 11:47:58.473001 22388622222912 subprocess_utils.py:97] Finished Hmmbuild in 0.533 seconds
I1127 11:47:58.476358 22388622222912 subprocess_utils.py:68] Launching subprocess "/hmmer/bin/hmmsearch --noali --cpu 8 --F1 0.1 --F2 0.1 --F3 0.1 -E 100 --incE 100 --domE 100 --incdomE 100 -A /tmp/tmpv0woh752/output.sto /tmp/tmpv0woh752/query.hmm /root/public_databases/pdb_seqres_2022_09_28.fasta"
I1127 11:48:08.666570 22388622222912 subprocess_utils.py:97] Finished Hmmsearch in 10.190 seconds
I1127 12:15:38.529495 22391989208192 pipeline.py:108] Deduplicating MSAs and getting protein templates took 1660.74 seconds for sequence GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 12:15:38.531214 22391989208192 pipeline.py:115] Filtering protein templates for sequence GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 12:15:38.536204 22391989208192 pipeline.py:124] Filtering protein templates took 0.00 seconds for sequence GMRESYANENQFGFKTINSDIHKIVIVGGYGKLGGLFARYLRASGYPISILDREDWAVAESILANADVVIVSVPINLTLETIERLKPYLTENMLLADLTSVKREPLAKMLEVHTGAVLGLHPMFGADIASMAKQVVVRCDGRFPERYEWLLEQIQIWGAKIYQTNATEHDHNMTYIQALRHFSTFANGLHLSKQPINLANLLALSSPIYRLELAMIGRLFAQDAELYADIIMDKSENLAVIETLKQTYDEALTFFENNDRQGFIDAFHKVRDWFGDYSEQFLKESRQLLQQANDLKQG
I1127 12:15:46.089067 22391989208192 pipeline.py:160] processing 2PV7, random_seed=1
Running AlphaFold 3. Please note that standard AlphaFold 3 model parameters are
only available under terms of use provided at
https://github.com/google-deepmind/alphafold3/blob/main/WEIGHTS_TERMS_OF_USE.md.
If you do not agree to these terms and are using AlphaFold 3 derived model
parameters, cancel execution of AlphaFold 3 inference with CTRL-C, and do not
use the model parameters.
Found local devices: [CudaDevice(id=0)]
Building model from scratch...
Processing 1 fold inputs.
Processing fold input 2PV7
Checking we can load the model parameters...
Running data pipeline...
Processing chain A
Processing chain A took 2574.40 seconds
Processing chain B
Processing chain B took 0.10 seconds
Output directory: /root/af_output/2pv7
Writing model input JSON to /root/af_output/2pv7
Predicting 3D structure for 2PV7 for seed(s) (1,)...
Featurising data for seeds (1,)...
Featurising 2PV7 with rng_seed 1.
Featurising 2PV7 with rng_seed 1 took 10.86 seconds.
Featurising data for seeds (1,) took  18.03 seconds.
Running model inference for seed 1...
Running model inference for seed 1 took  288.41 seconds.
Extracting output structures (one per sample) for seed 1...
Extracting output structures (one per sample) for seed 1 took  1.21 seconds.
Running model inference and extracting output structures for seed 1 took  289.62 seconds.
Running model inference and extracting output structures for seeds (1,) took  289.62 seconds.
Writing outputs for 2PV7 for seed(s) (1,)...
Done processing fold input 2PV7.
Done processing 1 fold inputs.

[linuxfold]

What GPU are you using? See issue #59

[evans-ZH]

oh, thanks, I use Tesla V100 (capability 7.0). It seems that capability 7.x isn't supported now.

[linuxfold]

oh, thanks, I use Tesla V100 (capability 7.0). It seems that capability 7.x isn't supported now.

Apparently they just issued a fix in #59 :

Hi all, in 781d8d0 we've added instructions to workaround the issue with CUDA Capability 7.x GPUs.

The workaround is to set the environment variable XLA_FLAGS="--xla_disable_hlo_passes=custom-kernel-fusion-rewriter" (see Dockerfile, docs/performance.md, docs/known_issues.md)

We've tested this fix on V100 to confirm it works, but have not tested on other CUDA Capability 7.x GPUs. Can you please give this workaround a try and let us know if it resolves the issue for you?

[evans-ZH]

Great, thanks!