# Average/Representative Structure from Trajectory

**URL:** <https://gromacs.bioexcel.eu/t/average-representative-structure-from-trajectory/2836>\
**Category:** User discussions\
**Created:** [September 19, 2021, 7:25pm UTC](https://gromacs.bioexcel.eu/t/average-representative-structure-from-trajectory/2836 "2021-09-19T19:25:40Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![joshhayden](https://avatars.discourse-cdn.com/v4/letter/j/aeb1de/32.png) [@joshhayden](https://gromacs.bioexcel.eu/u/joshhayden)\
**Post date:** [September 19, 2021, 7:25pm UTC](https://gromacs.bioexcel.eu/t/average-representative-structure-from-trajectory/2836/1 "2021-09-19T19:25:40Z")

</div>

GROMACS version: 2020.4  
GROMACS modification: No

Hi everyone,

Does anyone know of a simple command to produce a coordinate file of the average coordinates from the MD simulation? Or perhaps a way to produce a coordinate file of a frame from the simulation that is the most representative frame from the entire trajectory?

I am ultimately looking to overlay this representative structure on a PDB file to see if any large structural perturbations have occurred from the introduction of a single amino acid substitution. I wasn’t sure if it would be best to use the average coordinates, representative frame, or final frame from the 200-ns simulation I had run.

Any tips or guidance is appreciated!

Josh

---

<div class="post-metadata">

**Author:** ![sayantan0510](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/sayantan0510/32/1344_2.png) [@sayantan0510](https://gromacs.bioexcel.eu/u/sayantan0510)\
**Post date:** [September 21, 2021, 4:50pm UTC](https://gromacs.bioexcel.eu/t/average-representative-structure-from-trajectory/2836/3 "2021-09-21T16:50:40Z")

</div>

Hi,

I was thinking if you can remove the rotation and translation around the centre of mass of the molecule of interest throughout the trajectory…then you can write a simple ‘awk’ script to find out the average coordinate of each atom. Of course, you need to ensure that the conformation of the molecule has converged.

Best,  
Sayantan

---

<div class="post-metadata">

**Author:** ![joshhayden](https://avatars.discourse-cdn.com/v4/letter/j/aeb1de/32.png) [@joshhayden](https://gromacs.bioexcel.eu/u/joshhayden)\
**Post date:** [September 30, 2021, 1:21pm UTC](https://gromacs.bioexcel.eu/t/average-representative-structure-from-trajectory/2836/4 "2021-09-30T13:21:28Z")

</div>

Are you asking if the RMSD of the mutated protein is converging as in leveling off?

---

<div class="post-metadata">

**Author:** ![Anna13](https://avatars.discourse-cdn.com/v4/letter/a/e47774/32.png) [@Anna13](https://gromacs.bioexcel.eu/u/Anna13)\
**Post date:** [September 30, 2021, 3:30pm UTC](https://gromacs.bioexcel.eu/t/average-representative-structure-from-trajectory/2836/5 "2021-09-30T15:30:00Z")

</div>

Hi,  
I don’t know if I got your point but …  
In order to obtain the most representative conformation of a system you could perform a cluster analysis of the coordinates of your trajectory (gmx cluster)  
I suppose you are working with a protein…  
Depending on the selection you do, gmx cluster will perform the analysis on coordinates of the whole protein or just on coord. of backbone or C alpha etc…  
Be careful because the calculation is heavy, so be sure to not perform it for every ps of simulation…

good luck and let us know!