After you are familiar with using gromacs on a regular simulation, let's try temperature replica exchange simulation.

Here is a tutorial I used.

1)topology generation:

pdb2gmx -f 1e0q.pdb(default output: conf.gro, topol.top and posre.itp )

2)energy minimization:

grompp -f min.mdp -c conf.gro -p topol.top -o min.tpr

mdrun -v -deffnm min

Use http://folding.bmc.uu.se/remd/ to choose the temperature of the replicas depending on Tmin, Tmax and the number of replicas, N.

Temperature (K) μ (kJ/mol) σ (kJ/mol) μ12 (kJ/mol) σ12 (kJ/mol) P12
1 300.00 -5358 60.25
2 307.46 -5329 60.90 28.6 85.67 0.7000
3 315.14 -5299 61.57 29.5 86.60 0.7000
4 323.04 -5269 62.25 30.3 87.56 0.7001
5 331.18 -5238 62.96 31.2 88.54 0.7000
6 339.55 -5206 63.69 32.1 89.56 0.7000
7 348.16 -5173 64.44 33.1 90.60 0.7000
8 357.03 -5139 65.21 34.0 91.68 0.7001
9 366.16 -5104 66.01 35.1 92.79 0.6999
10 375.56 -5068 66.82 36.1 93.93 0.7000
11 385.24 -5030 67.67 37.1 95.10 0.7000
12 395.20 -4992 68.53 38.2 96.31 0.7001
13 405.32 -4953 69.41 39.3 97.55 0.7000

Construct a series of N input files (.tpr) from the minimised (in case of explicit solvent: equilibrated) structures, using different .mdp files to generate the different .tpr files.

 

for i in $(seq 0 12); do grompp -f sd_$i.mdp -c min.gro -p topol.top -o sd_$i.tpr; done

mpirun -np 13 mdrun_mpi -s sd_.tpr -multi 13 -replex 1000 -reseed 175320

Post-Processing:
GROMACS writes trajectories per temperature (temperature replicas), which are not continuous with
respect to coordinates (as the coordinates get exchanged). You can use the PERL script demux.pl that
will read your *.log files (you can concatenate them beforehand) and produce a few output files.
Example:
~$ cat md*.log >remd.log
~$ demux.pl remd.log

which produces the file replica_index.xvg. This file can be passed to trjcat, along with the original trajectory files, in order to produce continuous coordinate trajectories (coordinate replicas):
~$ trjcat -f *.trr -demux replica_index.xvg