cmap calculation fails for 2nd component - heterotypic system

[SimonRuedisser]

Test system:

Tau35_TIA1 from the CALVADOS examples
https://github.com/KULL-Centre/CALVADOS/tree/main/examples/slab_IDR_MDP

The calculation of cmap for TIA1 fails (see run.py and error below).

The calculation of cmap for Tau35 and between Tau35 and TIA1 completes without errors.

last few lines of run.py (request cmap for TIA1) to request computation of cmap for TIA1:

....
calculate contact map between name_2 and name_2
chainid_dict = dict(TIA1 = (100,199))
calc_contact_map(path="/nfs/nas22/fs2202/biol_bc_allain_1/projects/hnRNP-40S_gp/CALVADOS/examples/test_simon/Tau35_TIA1",sysname="Tau35_TIA1",output_path="data",chainid_dict=chainid_dict,is_slab=True)

slab_IDR.err output:

File "/nfs/nas22/fs2202/biol_bc_allain_1/projects/hnRNP-40S_gp/CALVADOS/examples/test_simon/Tau35_TIA1/run.py", line 42, in
calc_contact_map(path="/nfs/nas22/fs2202/biol_bc_allain_1/projects/hnRNP-40S_gp/CALVADOS/examples/test_simon/Tau35_TIA1",sysname="Tau35_TIA1",output_path="data",chainid_dict=chainid_dict,is_slab=True)
File "/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/calvados/analysis.py", line 965, in calc_contact_map
cmap += (.5-.5*np.tanh((d-1.)/.3)).reshape(mask_frames.sum(),
ValueError: cannot reshape array of size 5504812500 into shape (785,275,newaxis,275)

[sobuelow]

Hi Simon, thanks for raising this. Indeed, there is a bug in the slab contact analysis if the chains of the first mentioned component do not start with id=0.
The changes in the PR should fix this issue.
Note that the analysis with is_slab=True only calculates the contacts between the most central chain (in z dimension) of the first-named component in chainid_dict and all the surrounding chains of the other component. The most central chain of component 1 is reevaluated for each frame in the trajectory. If you intend to calculate all chains to all chains contacts regardless of where they are in the slab, you should set is_slab=False.

[SimonRuedisser]

Hi Sören,
many thanks for the quick response. I did download the new analysis.py and started a few runs with the example from above. However, I do run in out-of-memory problems (sbatch oom_kill event). I did request 4 CPUs with 64 Gbyte memory each on the cluster for the analysis only.
Any help is very much appreciated!
best, Simon

slab_IDR.err:

/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/calvados/analysis.py:638: RuntimeWarning: Mean of empty slice.
cden = hm[bool_dense].mean() # average dense concentration
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/numpy/core/_methods.py:192: RuntimeWarning: invalid value encountered in scalar divide
ret = ret.dtype.type(ret / rcount)
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/calvados/analysis.py:639: RuntimeWarning: Mean of empty slice.
cdil = hm[bool_dilute].mean() # average dilute concentration
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/calvados/analysis.py:643: RuntimeWarning: Mean of empty slice.
denarray = np.apply_along_axis(lambda a: a[bool_dense].mean(), 1, h)
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/calvados/analysis.py:644: RuntimeWarning: Mean of empty slice.
dilarray = np.apply_along_axis(lambda a: a[bool_dilute].mean(), 1, h) # concentration in range [bool_dilute]
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/calvados/analysis.py:638: RuntimeWarning: Mean of empty slice.
cden = hm[bool_dense].mean() # average dense concentration
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/numpy/core/_methods.py:192: RuntimeWarning: invalid value encountered in scalar divide
ret = ret.dtype.type(ret / rcount)
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/calvados/analysis.py:639: RuntimeWarning: Mean of empty slice.
cdil = hm[bool_dilute].mean() # average dilute concentration
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/calvados/analysis.py:643: RuntimeWarning: Mean of empty slice.
denarray = np.apply_along_axis(lambda a: a[bool_dense].mean(), 1, h)
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/calvados/analysis.py:644: RuntimeWarning: Mean of empty slice.
dilarray = np.apply_along_axis(lambda a: a[bool_dilute].mean(), 1, h) # concentration in range [bool_dilute]
/cluster/home/gpalfy/miniconda3/envs/calvados/lib/python3.10/site-packages/pandas/core/internals/blocks.py:2540: RuntimeWarning: invalid value encountered in cast
values = values.astype(str)
Killed
[2026-02-16T16:00:16.992+01:00] error: Detected 1 oom_kill event in StepId=57322853.batch. Some of the step tasks have been OOM Killed.

[sobuelow]

Hi, thanks again. I have found and at least temporarily fixed the memory issue in the above PR #48. Could you please try it out? It's quite slow, but I'm hesitating to re-write larger sections to speed it up as I did not write that part of the analysis myself and it could become a bit of a rabbit hole...

[SimonRuedisser]

Hi Sören,
many thanks for fixing the issue. I have tested it, it does now calculate the cmap for both homotypic and the heterotypic system. Best, Simon

[sobuelow]

Great to hear!