I do not know which one is the best, but I can tell you some tools that I used.

For protein structure prediction, I use a web server called swiss-model, and I think another one called Robetta is also good.

For refinement, especially in cryo-EM field, I think rosetta, phenix, COOT, MDFF, and Direx are all good.

For validation I will use the validation module in PDB and MD simulation, I have no good idea.

Hope it will help.

Robert, if you are interested only in the 3D protein structure then Swiss-model server is the easiest and best way. The limitation with this automated modeling comes, if you need to model some ligand molecules, or lets say water. In that case I would suggest to use Modeller, where you can build/predict accordingly.

If you are dealing with a protein sequence, that does not have good template from BLAST or other alignment search, then go for other methods e.g threading, and you can find many online servers. For your references, I-Tasser, Robetta, Ev-Fold, Phyree etc.

After generating some models, you need to choose the best one for further analysis therefore submit your structure to RAMPAGE server and ProSa servers where you can find the steriochemichal and energy landscapes of your models. Lastly, try some minimization too, if needed.

I guess you want to transfer that structure into MD studies, well, I have used NAMD therefore I would suggest this one. Other MD packages may be better or worse, i have no idea about that.

Good Luck

In addition to Ashfaq Ahmad's comments, I would suggest gromacs and Desmond for MD simulation. Both are open source. Gromacs is widely used and there are good tutorials available. Desmond comes with Schrodinger Maestro interface. Moreover, ligand preparation can be done more easily in Schrodinger Maestro.

For MD simulation , you can use GROMACS. Very good software.

Thanks Aditi Rathee and Uttam Pal and Ashfaq Ahmad and Mingliang Jin for your answers.

I am grateful