Gmx dssp gives wrong secondary structure results

User discussions
1

GROMACS version: 2024.1
GROMACS modification: No

Hi everyone,

I’m using the current GROMACS version that contained a new “dssp” command and removed the old “do_dssp”. However, when I was trying to use the following command:

gmx dssp -s md_start.pdb -f md_aligned.xtc -o …/analysis/dssp.dat -num …/analysis/dssp_num.xvg

the command only gives a “.dat” output file containing “~”, “S”, and “P”, which is entirely unphysical because my protein obviously contained alpha helices and beta sheets (with VMD).

For example, MDTraj gives this assignment:

~~~~~HHHHHHHHHTT~~~SS~~HHHHHHHHHHHHHHTT~B~SSS~SEEE~TTSHHHHHHT~SEEETT~HHHHHHHHHHHS~~~HHHHHHHHHHS~~~~

while gmx dssp gives this:

~~~~~SSSSSSSSSS~~~~SS~~SSSSSSSSSSSSSSS~~S~SSS~S~~~~SSSSSSSSSS~S~~~SS~~SSSSSSSSSSS~~=~~SSSSSSSSSSPP~~

This mismatch is so incomprehensible… Am I calling the gmx dssp the wrong way? Or is it a problem with gmx dssp?

Thank you all.

2

Hi,

Minor mismatches between different tools are expected, but what you describe does not look reasonable.

Would you mind sharing the input files? Ideally, check if the problem can be reproduced with a single frame instead of the full trajectory.

Pinging @alexxy.

3

Hi,

We did tests with some set of randomly selected proteins from pdb, and checked that output of gmx dssp and original DSSPv4 was the same. So I agreed with @al42and that to check if its issue with gmx dssp we need example/test files which have differences in assignment of secondary structures between two tools.

4

Hello,

Thanks for your comments. I cannot disclose the pdb and the trajectory files in my original post, but I have another input pdb and the xtc trajectory available in this link that has the same problem: gmx_test_dssp.zip - Google Drive

And the gmx dssp result for one single frame (for the pdb only, with no trajectory input) was:

~~~S~~SSSSS~SSSSSSSSSSSSSS~SSS~~~~~~~~SS~~SS~~~S~~~~~~~SS~~SSSSSSSSSSSS~~~SSS~~~~~SSSS~~~~SS~~SSSSSSSS~~SSS~~~~~~~SS~~~S~~~~~

which was totally unreasonable (when compared with VMD visualization).

Please feel free to check out. Thank you very much.

5

You’ll have to use --hmode dssp --clear options if you’re going to use .pdb extension(s) as input. Otherwise, hydrogen atoms will not be recognized.