Replies: 7 comments 7 replies
-
|
The snugdock.static.linuxgccrelease application is using just one core. I
don't know for certain whether snugdock supports MPI parallelization, but
it probably does, which means you could get a (roughly, at best) 32 fold
speedup by using the MPI build and running on all 32 cores.
…On Mon, Sep 2, 2024, 1:55 AM Wanggang ***@***.***> wrote:
The antibody-antigen docking with Snugdock was done on a Ubuntu 22.04 LTS
with a 32 CPU with the following command: snugdock.static.linuxgccrelease
-dock_pert 3 8 -partners LH_A -spin -h3_filter false -mute core.io.database
-loops:refine_outer_cycles 2 -loops:max_inner_cycles 20 -detect_disulf
false -antibody:check_cdr_chainbreaks false -nstruct 1000 -s
$prepacked_complex.pdb. I found that the docking speed is too slow. The
estimated time per run is about ten days. By the way, flag "-nstruct 1000"
is the recommended setting (
https://docs.rosettacommons.org/docs/latest/application_documentation/antibody/snugdock).
Could you tell me how to deal with this issue? Thank you!
—
Reply to this email directly, view it on GitHub
<#140>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABACZUFRZTJOM5RUP4D4MOTZUOSQVAVCNFSM6AAAAABNPISTJOVHI2DSMVQWIX3LMV43ERDJONRXK43TNFXW4OZXGEZDEMJRGQ>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
According to your advice, I have downloaded the source file of Rosetta 3.14 and compiled the MPI version of Rosetta using the following command: python ./scons.py -j 30 mode=release bin extras=mpi. Finally, the generated bin folder contains 456 items (see below). I am not sure whether I have successfully compiled the MPI version. Becasure these 456 items do not mention "MPI" characters in their names. Moreover, the compiling process in terminal mentioned a error "g++: warning: switch '-ffor-scope' is no longer supported". When I run Snugdock using either snugdock.default.linuxgccrelease or snugdock.linuxgccrelease, the computation speed is still very slow. Could you guide me how to solve it? Thank you! 456 items in bin folder compiling is complete: |
Beta Was this translation helpful? Give feedback.
-
Hmm. Looks like something's going wrong with the MPI compilation, if you're not getting files of the form |
Beta Was this translation helpful? Give feedback.
-
|
Thank you for your reply! Where is the compliation log? I have uploaded the final running log shown in terminal. |
Beta Was this translation helpful? Give feedback.
-
Oh, weird! It looks like it's just doing a default compilation. Hmm. Try putting "bin" last in your compilation command: |
Beta Was this translation helpful? Give feedback.
-
|
(The individual compilation commands should invoke the MPI C++ compiler -- usually |
Beta Was this translation helpful? Give feedback.
-
|
Your advice is excellent. I have successfully installed the mpi version of rosetta. I really appreciated your help!
发件人："Vikram K. Mulligan" ***@***.***>
发送日期：2024-09-05 13:54:56
收件人：RosettaCommons/rosetta ***@***.***>
抄送人：Wanggang ***@***.***>,Author ***@***.***>
主题：Re: [RosettaCommons/rosetta] The docking speed of antibody-antigen docking with Snugdock is too slow. (Discussion #140)
Thank you for your reply! Where is the compliation log? I have uploaded the final running log shown in terminal.
Oh, weird! It looks like it's just doing a default compilation.
Hmm. Try putting "bin" last in your compilation command:
./scons.py -j 30 mode=release extras=mpi bin
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you authored the thread.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
Ok. I got it. Thank you very much!
发件人："Vikram K. Mulligan" ***@***.***>
发送日期：2024-09-05 13:56:00
收件人：RosettaCommons/rosetta ***@***.***>
抄送人：Wanggang ***@***.***>,Author ***@***.***>
主题：Re: [RosettaCommons/rosetta] The docking speed of antibody-antigen docking with Snugdock is too slow. (Discussion #140)
(The individual compilation commands should invoke the MPI C++ compiler -- usually mpicxx or mpic++ or some such.)
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you authored the thread.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
OK, great! The next step is running it. I double-checked the source code, and the If this is still too slow, and if you have access to a large computing cluster, you can also run on many cores on many nodes across the cluster. You'll have to refer to your cluster's documentation for the syntax, however. A few notes:
|
Beta Was this translation helpful? Give feedback.
-
|
One last thing to be aware of: many flavours of MPI do not treat hardware threads as separate cores. Many modern CPUs advertise, for example, 64 threads, but have only 32 physical cores, relying on hyper-threading to allow performance somewhat close to what you'd get with 64 cores for some tasks. Be sure to either (a) launch as many processes as you have physical cores, or (b) use the |
Beta Was this translation helpful? Give feedback.
-
|
OK, thank you very much for your further advice and guidance. I have tested mpirun command (at least 20 physical cores available for my 32 CPU system). The computation speed has increased a lot. Here, I would like to ask an question regarding result evaluation. The resulting score.sc file contained many evaluation parameters for each conformation of the docked antiboby-antigen complex. Then, how to evaluate the binding affinity between antibody and antigen using the score.sc file? |
Beta Was this translation helpful? Give feedback.
-
|
If the scorefile has a column like That said, empirically, this value is only weakly correlated with binding affinity (i.e., very good values are somewhat more likely to be strong binders than very bad values). Generalizable protein-protein binding affinity prediction across different antigens (or, indeed, across different antibody germlines) remains unsolved. |
Beta Was this translation helpful? Give feedback.
-
|
@everyday847 Ok. Thank you for your kind advice! |
Beta Was this translation helpful? Give feedback.
-
|
Marking this as "closed", as I hope your question has been answered. If that assessment is incorrect, feel free to re-open or open up a new discussion with more details. |
Beta Was this translation helpful? Give feedback.
-
The antibody-antigen docking with Snugdock was done on a Ubuntu 22.04 LTS with a 32 CPU with the following command: snugdock.static.linuxgccrelease -dock_pert 3 8 -partners LH_A -spin -h3_filter false -mute core.io.database -loops:refine_outer_cycles 2 -loops:max_inner_cycles 20 -detect_disulf false -antibody:check_cdr_chainbreaks false -nstruct 1000 -s $prepacked_complex.pdb. I found that the docking speed is too slow. The estimated time per run is about ten days. By the way, flag "-nstruct 1000" is the recommended setting (https://docs.rosettacommons.org/docs/latest/application_documentation/antibody/snugdock). Could you tell me how to deal with this issue? Thank you!
Beta Was this translation helpful? Give feedback.
All reactions