'BioSimSpace._SireWrappers.System' can only contain unique molecules. about biosimspace HOT 8 CLOSED

This error occurs in both the 2022 and 2023 versions.

from biosimspace.

I've not had a chance to test, but assume the error is occurring since the objects have the same MolNum. To get a unique number, you could create a copy of one, e.g.

print(morph + pro.copy())

I guess the issue here is that MolNum identifiers are unique to a Sire session, so there can be a clash if you pickle/unpickle and object. I think we will need to implement a copy of unpickling, or similar, although this could lead to other issues.

from biosimspace.

Hi @lohedges , many thanks for your quick reply!

I've tried using the .copy() function, but the MolNum remains unchanged.

from biosimspace.

Ah, I think it only works with the .copy done at the Molecule level, not at the System. Try extracting the molecules and doing a copy of one before recombining to make a new system.

from biosimspace.

Thanks, pro[0].copy() works! 🎉 🎉

from biosimspace.

Great. I'll update the System copy method to copy each of the molecules in it, then recombine them.

from biosimspace.

Just a quick update on this, since we might need to have a rethink on how we constrain and compare molecules in BioSimSpace. In Sire, there is the concept of a molecule number and version. This means you can have different versions of the same molecule number, i.e. if you edit it. The combined version and number is guaranteed to be unique.

For BioSimSpace, I have needed to place some additional restrictions on molecule (and residue and atom) numbering within a System due to the requirements of SOMD, i.e. it assumes unique and ascending numbering, even though this isn't a requirement of SIre itself. For example, SOMD completely re-orders the atoms in a system on load, so we need to use some unique identifiers, e.g. atom number, to re-map updated coordinates from a trajectory file.

For the above, doing a .copy() doesn't really make sense in terms of the behaviour of Sire, so we are thinking of changing this to .renumber(). We had discussed implicitly doing this on unpickling, but this could cause other issues, e.g. if you wanted to unpickle a System and Molecule, then check that the Molecule was in the System.

When we eventually re-write SOMD I will make sure that it is designed in a way that doesn't require invalid assumptions about a Sire system. This will hopefully avoid issues like this.

Just to note that we realised that your issue could also be resolved by changing the order of the read and pickle in your second script, i.e. if you unpickle first, then load the perturbable system, i.e.:

import pickle
import BioSimSpace as BSS
with open('pro.pkl', 'rb') as f:
    pro = pickle.load(f)
morph = BSS.IO.readPerturbableSystem('morph0.prm7', 'morph0.rst7', 'morph1.prm7', 'morph1.rst7')
print(morph + pro)

This is because the molecule in morph is correctly assigned a unique number on read.

from biosimspace.

Thank @lohedges for the detailed information 🤗

Just to note that we realised that your issue could also be resolved by changing the order of the read and pickle in your second script, i.e. if you unpickle first, then load the perturbable system, i.e.:

Thanks! that's good to know.

from biosimspace.