There are many rules to determine drug-likeliness of a particular chemical compound. Lipinski's rule of 5 is one of them with few postulates one of them being logP i.e water-octanol partition coefficient.
I suppose you could in theory use any non-polar solvent that is immiscible with water. However, the properties of 1-octanol are thought to resemble those of lipid bilayer membranes rather well (i.e., the distribution of chemicals/drugs into 1-octanol simulates their ability to passively diffuse across membranes in living systems). Until someone discovers a better solvent system, water/1-octanol will remain as the standard.