Virtual screening is a well-known approach in drug discovery projects to find new leads in virtual libraries of small molecules, but it is not clear which of the existing techniques is better.
Ligand-based tools are characterized by their speed but do not take into account the receptor while structure-based tools model ligand-receptor interactions but are more compute intensive.
It has been shown that in many projects simpler approaches work better, which might seem counterintuitive, but this is due to the noise introduced by complex models.
What do you think?