Virtual screening: an endless staircase?

Computational chemistry - in particular, virtual screening - can provide valuable contributions in hit- and lead-compound discovery. Numerous software tools have been developed for this purpose. However, despite the applicability of virtual screening technology being well established, it seems that there are relatively few examples of drug discovery projects in which virtual screening has been the key contributor. Has virtual screening reached its peak? If not, what aspects are limiting its potential at present, and how can significant progress be made in the future?