Contrasting results of CSP studies, showing the lattice energies of the generated structures against the packing coefficient (Gavezzotti, 1983), with the observed structures circled in red. (a) Only one structure is thermodynamically plausible. The experimental structure of isocaffeine (XVb) is clearly predicted. (b) Whilst the observed structure is the most stable, there are other structures on the crystal energy landscape requiring consideration as to whether they could be observed polymorphs. The crystal structures of GSK269984B (XI) are classified by their conformation, a torsion angle that strongly affects the gross conformation and the carboxylic acid conformation. Since the thermodynamically competitive structures have very different gross conformations, these crystal structures would be expected to be long-lived if they crystallized (Ismail et al., 2013). (c) There is a cluster of layer crystal structures around the global minimum, requiring analysis as to whether they could crystallize as distinct polymorphs or might lead to disorder. Both polymorphs of caffeine (XVa) are disordered layer structures. The crystal structures are classified according to their molecular stacking (Habgood, 2011).