We can only tell from the example data.
pelagic's regex is also unsafe. what if the line looks like

ATOM   3316  N   TYR B  24      97.479  18.805   2.990  1.00 36.32    
+  D    N
[download]

Then the alone "D" at the end will be replaced by "X". Intended??

I would use the substr()-solution above. My regex was just meant as an optimization of perlsen's one, which is unneccesary complicated.

Update:
Striked the sillyness about the second "D".
/me writes "Must read the question properly" 200 times.