The data set should contain structures and have had a scoring profile applied. This enables molecules to be selected that have the best possible balance between performance against the scoring profile and structural diversity. In this case the user can determine the bias between 'Rank' and 'Diversity'.




