Problem (Bug?) in RDF calculation for Group-Based RDFs in 2021
There appears to be a problem with calculation of RDFs between center of masses between globular protein molecules. When calculating (using whole_mol_com selections) the C.O.M. RDF between the globular proteins (radius of gyration around 2-2.5 nm), code version 2021.1 produces spurious peaks in the RDF below 2 nm, whereas version 2020.1 (standard ubuntu 20.4) does not produce these spurious peaks. Visual trajectory inspection proves that the protein molecules do NOT denature, thus, no spurious peaks should occur.
Command Line for the RDF calculation:
gmx rdf -pbc -rmpbc -f ../../protein_mod50.xtc -s topol.tpr -n index.ndx \
-selrpos whole_mol_com -seltype whole_mol_com -norm rdf -o rdf.xvg \
-b 6000 -e 8000 -rmax 16 -bin 0.1
Further info: Similar (but less pronounced) spurious problems occur when using the res_com selection scheme.
Additional info: this seems not to be a problem of the method to generate the topol.tpr/index.ndx files, since these were created both by the 2020 and the 2021 gromacs version, with identical bad results when running gmx rdf from the 2021 code suite. I also noted that, when calculating the RDF for short 2ns segments of the long trajectory that sometimes these spurious peaks occur, sometimes not, which might be an indication that this is an initialization problem?
I append a comparison of the RDFs from two different time intervals of the trajectory for demonstration. For better visibility I shifted the difference between 2021 and 2020 down by 0.1 units and the 2021 version up by 0.1 units. The orange 2020 curve is physically reasonable.
Thanks for looking into this.
Eckhard