If your model is linear, then there is no reason to explicitly run both positive and negative g loads.
BUT, if you are looking for restraint loads, it might be simpler to run both to get the range of loads; for example, 1) operating plus g in X and operating minus g in X.
Or:
L1:W+T1+P1 (OPE)
L2:W+P1 (SUS)
L3:U1 (OCC)
L4:W+T1+P1+U1 (OPE)
L5:W+T1+P1-U1 (OPE)
L6:L1-L2 (EXP) [ALGEBRAIC]
L7:L2+L3 (OCC) [SCALAR]
Where:
use L1,L2,L4&L5 for strucutal evaluation (displacements & restraint loads)
use L2,L6&L7 for stress evaluation

If you have nonlinear restraints in your model, then more work is required.