Hi Kio Amara . Implementing molecular dynamics in python would be a lot of work, and I wouldn't recommend it as there are already many reliable and well-developed packages such as the one you just mentioned (Gromacs, Amber, etc.).

In terms of performance, I think the current packages implemented in C++ or fortran are much faster than python (again because they are being developed for many years now.)

You have to use the PDB file, and come up with an input file, in which you can change the temperature and not in the PDB file.

You can use python in post-processing of your results; For that also have a look at packages such as MDAnalysis or ASE.

Another thing you can use python for, is to automatise generation or modification of input files for your simulations, or their submission to a computing node.

GROMACS has a very robust python interface (https://manual.gromacs.org/current/gmxapi/userguide/usage.html).

If LAMMPS is also an option for your work you can find this alternative (https://lammps.sandia.gov/doc/Python_head.html).

I have been using Python to implement polymers lattices for MD, however, I am gradually migrating to Julia due to higher computational performance.