You could also use a Python library such as MDAnalysis and write a script that does exactly what you want. It could look something like:

```python

import MDAnalysis as mda

import numpy as np

u = mda.Universe('topology.gro', 'trajectory.xtc')

# select first set of residues; in this example resids 21 - 27

residues1 = u.select_atoms('protein and resid 21:27').residues

# select second set of residues; in this example resids 38 - 42

residues2 = u.select_atoms('protein and resid 38:42').residues

# calculate distances between each CA in first group with each CA in second group

from MDAnalysis.lib.distance import distance_array

dists = distance_array(residues1.select_atoms('name CA').positions, residues2.select_atoms('name CA').positions)

# or, calculate distances between the COM of each residue in first group with COM of each residue in second group

residues1_com = np.array([res.center_of_mass() for res in residues1])

residues2_com = np.array([res.center_of_mass() for res in residues2])

dists = distance_array(residues1_com, residues2_com)

```

Unfortunately ResearchGate doesn't have a decent way of posting code in the Q&A box, so you can copy and paste this into a text editor to make it more readable. 

You still have to decide what distance (CA, COM, something else) is most relevant to the biological question you wish to ask.

http://mdanalysis.org/

You should clarify your goal. Do you want to calculate minimal distance between residues? Then g_mindist is what you need.

Tanks for your prompt answer. I want to have a residue-residue distance matrices for two series of amino acids and in last frame of my simulation.

I've used this command : g_mindist -s comp2.tpr -f md.nojump.fit2.xtc -n dist_martics.ndx -od mindist1.xvg -or mindistres1.xvg -o inter3.xvg -b 199999 -e 200000

but the output files were not what I wanted.

truly what I need is distances between any amino acids of group 1 from any amino acids of group 2. (not center of mass of the whole group )

Your question is somewhat poorly defined. How would you define the distance between two residues if not the center-of-mass distance between them? To me, what it sounds like you want is, in fact, a loop of calls to gmx distance between index groups, each of which defines an individual residue. You can then take an average of each time series and build the distance matrix you want. Simply put, there is no single tool that will output exactly what you want; you will have to take several steps and perhaps do some external scripting to build this matrix.

As answered by others, you should define ur exact goal. Two amino-acids have a number of atoms. So, looking for a potential distance either to calculate minimal distance from their COM or a better approach is to measure the CA-CA distance. In case you are much focused on the importance of their side-chains as per your biological significance, you should proceed sequentially from graphical/diatance related constraint measurements to identify possibility of bond formation or inter or intra-residual interactions. So, as amino acids are group of atoms, logically we will calculate the mean distance of atom specific distances. U can refer to some NMR books or databases to understand the distance specific amino acid nomenclature. 

If you need to make a matrix u can simply design a small script like this specified index file 

Here group 24, 25 are CA  and 26,27 are N atoms of two residues

Design ur own to make a matrix .

#!/usr/bin

path=/user/path_name

cd $path

g_dist_d -f ../md.trr -s ../md.tpr -n index.ndx -o atomCA-CA.xvg

I really appreciate your help.I have made my self clear by editing my response above.

so I will try to design some scripts for my job.