Actually its not a common problem when you obtain your structure from x ray crystallographic results.In this situation with considering minimization failure I suggest docking to obtain initial structure. note that docking result must be compared with crystallographic structure (cluster selection based on crystallographic structure).
The clach is between two H atoms . While preparing files to undergo MD simulations, Hs of protein and ligand were built separatly and then two files were merged together. Due to that, the clach was appeared. But when adding H dirctly to the structure, no clashes were appeared. I don't know why the minimization didn't remove this clash.
N.B: I'm trying to minimize the system using steep and cg in GROMACS package .
First, try to do minimization with the very-very small time step (if one fails then try lower than this) it will help system to recover from overlapping of atoms and then do your normal minimization.