//TLDR

Problem: On one of my recent projects, the ligand appeared intact during sorting stage but distorted after converting the stream file.

Cause: This is caused by long ligand names, whihc distorts how the charmm2gmx script parses the stream file.

Solution: Maintain ligand/residue name character count to 4 or below.

Detailed explanation

Parameterization of a ligand is a crucial step in molecular dynamics simulations.The detailed tutorial for this process is provided by Dr JA Lemkul on his website. The tutorial is very detailed and provides a step-by-step guide to parameterize a ligand.

The steps involve

  1. Conversion of the ligand file from .pdb to .mol2 format. I generally use Open Babel for this purpose.

  2. In the second step, change the ligand name in the .mol2 file to a uniform notation. The first change that needs to be made is in the MOLECULE heading and also the residue names and numbers such that they are all the same. This is well illustrated in the tutorial.

  3. Sort the bonds using the sort_mol2_bonds.pl script.

  4. Upload the sorted .mol2 file to CGenFF server and download the CHARMM “stream” file (.str) file containing all of the topology information.

  5. Finally, generate topology using cgenff_charmm2gmx.py script.

On one of my recent projects, the ligand appeared intact in Step 3 but the output of Step 5 was distorted and mangled (Figure 1). I first thought that there was some problem with the server which mangled the output. I tried implementing other options, including one option here molecular dynamics - Alternative to CGenFF for generating large ligand topology - Matter Modeling Stack Exchange whihc was largely unsuccesful.

Figure 1

However, upon close examination of cgenff_charmm2gmx.py, I found that my ligand name was 5 characters long, which broke the PDB format, specifically how the python scrip parses the stream file. Upon truncating this name to 3 characters, the problem was fixed.

A more detailed discussion on this topic is available here:

  1. Gromacs Forums - https://gromacs.bioexcel.eu/t/problem-in-file-broken-structure-ini-pdb-generated-using-cgenff-for-proceeding-in-md-simulation/2926/6
  2. Matter Modeling Stack Exchange - https://mattermodeling.stackexchange.com/questions/8310/alternative-to-cgenff-for-generating-large-ligand-topology/8311#8311

Hope this helps! Do let me know if you have any questions.