# Atoms not completely removed by convert-tpr

**URL:** <https://gromacs.bioexcel.eu/t/atoms-not-completely-removed-by-convert-tpr/13526>\
**Category:** User discussions\
**Tags:** analysis-tools\
**Created:** [June 12, 2026, 1:21pm UTC](https://gromacs.bioexcel.eu/t/atoms-not-completely-removed-by-convert-tpr/13526 "2026-06-12T13:21:50Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Erik](https://avatars.discourse-cdn.com/v4/letter/e/b5ac83/32.png) [@Erik](https://gromacs.bioexcel.eu/u/Erik)\
**Post date:** [June 12, 2026, 1:21pm UTC](https://gromacs.bioexcel.eu/t/atoms-not-completely-removed-by-convert-tpr/13526/1 "2026-06-12T13:21:50Z")

</div>

GROMACS version: 2025.1  
GROMACS modification: No

I’m trying to perform analysis using only part of a molecule (Removing the ends of a nucleic acid strand). Because of how my downstream analysis pipeline works, I determined that the best way to handle this would be to use `trjconv` and `convert-tpr` with an index group to get xtc and tpx files excluding the unwanted residues. Modifying gro and xtc files with `trjconv` worked well, but when I generate a truncated tpx file with `convert-tpr`, it keeps the whole molecule topology, including the removed residues, resulting in incorrect atom types when loaded in downstream tools (MDAnalysis in this case). I unfortunately need to use the tpr file because there are more than 100k atoms in my system, so using the gro file as the topology for analysis results in indexing problems as my downstream analysis uses atom indexes. Any idea how to get a clean tpx file where the molecule topology is also truncated?

`convert-tpr`/`trjconv` commands:

```bash
gmx convert-tpr -s md.tpr -n noloops.ndx -o noloops.tpx
gmx trjconv -s md.tpr -f md.gro -n noloops.ndx -o noloops.gro

```

Looking at the gro file, it starts on the residue I want to start my analysis on:

```auto
$ head noloops.gro
RNA_chain_A
126618
    3U P 1 9.307 5.736 8.243 0.1655 -0.1133 -0.3574
    3U O1P 2 9.337 5.809 8.363 -0.6448 0.5286 -0.5350
    3U O2P 3 9.305 5.818 8.119 0.1602 0.0273 0.7769
    3U O5' 4 9.419 5.616 8.225 0.6936 -0.8166 -0.4415
    3U C5' 5 9.479 5.552 8.334 -0.3628 0.0396 -0.4869
    3U H5'1 6 9.402 5.507 8.397 0.5341 -2.7104 -1.2128
    3U H5'2 7 9.523 5.621 8.406 -2.3819 1.2063 -0.3085
    3U C4' 8 9.569 5.436 8.292 0.7754 -0.4253 0.1886

```

But when I load the tpx file with MDA:

```python
import MDAnalysis as mda
u = mda.Universe('noloops.tpx', 'noloops.gro') # Atom types are wrong
print(u.atoms[0].type, u.atoms[1].type, u.atoms[2].type)

u = mda.Universe('noloops.gro', 'noloops.gro') # atom types are right, but atom IDs reset at 99999
print(u.atoms[0].type, u.atoms[1].type, u.atoms[2].type)

u = mda.Universe('md.tpr', 'md.gro') # Full tpr as reference
print(u.atoms[0].type, u.atoms[1].type, u.atoms[2].type)

```

This prints:

```auto
OH HO CI
P O O
OH HO CI

```

As you can see, loading the truncated tpr file still results in atom types matching the removed 5’ end of the nucleic acid strand, rather than starting on nucleotide 3.

---

<div class="post-metadata">

**Author:** ![al42and](https://dub1.discourse-cdn.com/flex017/user_avatar/gromacs.bioexcel.eu/al42and/32/1393_2.png) [@al42and](https://gromacs.bioexcel.eu/u/al42and)\
**Post date:** [June 14, 2026, 11:58pm UTC](https://gromacs.bioexcel.eu/t/atoms-not-completely-removed-by-convert-tpr/13526/2 "2026-06-14T23:58:36Z")

</div>

Hi!

Looks like a bug, thanks for reporting! I opened [an issue](https://gitlab.com/gromacs/gromacs/-/work_items/5655), to be fixed in a 2026 patch release.

I cannot come up with any workarounds. Maybe manually reassigning `u.atoms.types` in the analysis script can work?

---

<div class="post-metadata">

**Author:** ![Erik](https://avatars.discourse-cdn.com/v4/letter/e/b5ac83/32.png) [@Erik](https://gromacs.bioexcel.eu/u/Erik)\
**Post date:** [June 15, 2026, 7:16am UTC](https://gromacs.bioexcel.eu/t/atoms-not-completely-removed-by-convert-tpr/13526/3 "2026-06-15T07:16:32Z")

</div>

Thanks for looking into it! I will figure out a work-around in the mean time.

---

<div class="post-metadata">

**Author:** ![rrsathe](https://avatars.discourse-cdn.com/v4/letter/r/a6a055/32.png) [@rrsathe](https://gromacs.bioexcel.eu/u/rrsathe)\
**Post date:** [July 8, 2026, 8:04am UTC](https://gromacs.bioexcel.eu/t/atoms-not-completely-removed-by-convert-tpr/13526/4 "2026-07-08T08:04:18Z")

</div>

Hi!

I tracked this down to convert-tpr not subsetting the associated atom metadata (such as atom types and optional PDB information) along with the atom records, resulting in the inconsistency you observed.

I’ve submitted a proposed fix with a regression test: **[MR ! 6106](https://gitlab.com/gromacs/gromacs/-/merge_requests/6106)**. If merged, this should resolve the issue in a future release.