If you are using standard AMBER residues, there is no need for prep. If you need to create new residues, you need to use prep to define them.
If so, should I begin with EDIT?
You should begin with LINK.
I would also like to combine a protein and a ligand which is not a protein. How can I simulate those compounds which are of different types? There might be a problem when I run prep.
You may need to define the ligand using prep. Normally one would specify the protein and ligand as separate molecules in LINK, which in any case will use an AMBER residue database for the protein and your own prep database for the ligand. See the triostin-linker demo.