Open tylerjereddy opened 8 months ago
Dear @tylerjereddy,
I apologize for the delayed answer, however, I was at a conference last week. In a nutshell, the parameters for branched Dextran that are in the input files to be used with the jupyter-notebook are a preliminary Dextran model, which minimally diverges from the one published in JCTC. In order to be safe the branched links were therefore removed from the current Martini 3 repository (until someone complains). Not an ideal solution but we're really short on people.
Restoring the branched links should be as straightforward as adding the following directive to the existing links and parametrizing the dihedral angles, which in the previous versions were not present.
[ link ]
; alpha 1->3
resname "DEX"
[ atoms ]
A { }
B {"replace": {"atype": "SP1r"}}
C { }
>A {"replace": {"atype": "SN6"}}
>B { }
>C { }
[ bonds ]
; i j funct length force constant
B >A 1 0.32 5500 {"group": "a13", "edge": "false"}
[ angles ]
; i j k funct angle force con.
C B >A 1 180 180 {"edge": "flase", "group": "CBA - a13"}
B >A >B 1 115 180 {"edge": "flase", "group": "BAB - a13"}
A B >A 1 108 180 {"edge": "flase", "group": "ABA - a13"}
B >A >C 1 85 162 {"edge": "flase", "group": "BAC - a13"}
[ dihedrals ]
A B >A >B 1 xxx
[ edges ]
B >A {"linktype": "a13"}
[ link ]
resname "DEX"
[ dihedrals ]
B +A +B ++A xxx {"version": "1", "group": "BABA - a13"}
If you want I can assist you in parametrizing them. I have the all-atom data and reference distributions. It just needs someone with a little time to optimize them. Let me know!
Working on latest
master
branch ofpolyply
.I was trying to roughly follow the tutorial at https://github.com/marrink-lab/polyply_regression_tests/blob/main/examples/liquid_liquid_phase_separation/PolyplyTutorial.ipynb
Then, I noticed that
dextran.martini3.ff
doesn't even have alpha 1->3 linkages anymore, they were deleted for some reason in 0a76a13 -- was that an accident? Perhaps a branched build regression test is needed there? It looks like you already have an unbranched dextran regression test atpolyply/tests/test_data/library_tests/martini3/DEX/polyply/command
.PEO polymer construction seems to work just fine for me locally, and so does unbranched dextran. For reproducer, this could be pared down quite a lot, but you'll get the idea--need to generate a branched JSON file with DEX 1-3 linkages/branches, then the failure to
gen_params
can be reproduced. I can help you produce a smaller regression test if you really want, but fixing the parameters in a MARTINI-3 compliant way is probably beyond the scope of my expertise, though you may just need to restore some deleted lines to fix it?