The "bank end", the "data" file, only needs relationships needed for calculations (or validation calculations). These relationship could be between tables, or self-relationships. You may want other relationships, for general navigation (optional, if needed for trouble-shooting). But the fewer the better.
The "front end" or "interface" file has most of the relationships. It may or may not have those mentioned above (often the relationships needed for calculations in the data file are also needed in the interface file, for other reasons).
There are usually very few scripts in the data file. I can think of two types. Those which must run with Full Access (which requires them to run in the data file),* and "import/update" script routines, to import data into a clean clone (which require no relationships).
*These are usually subscripts, called by the interface file. The option to [x] "Run with Full Access privileges" (lower left of scripts) does not cross files (for security reasons). So the parts of scripts requiring Full Access to data must run in the data file. No need to write the whole script there however; just call subscripts.