Conclusions

In this article, we have described novel methods for compound selection and library design. The compound selection method is a knowledge-based approach whereby weights are derived based on the discrimination between two classes of compounds used as training sets. The weights can then be used to score and rank compounds according to the extent to which they exhibit characteristics that are common to the compound classes used in training. For example, weights can be derived that discriminate between 'drug-like' and 'non-drug-like' compounds and can be used subsequently to order compounds for screening so that the compounds with 'drug-like' characteristics are screened first. Using the SELECT program for designing combinatorial libraries in product space, we have provided evidence, based on a number of different molecular descriptors and different diversity metrics, that product-based library design can lead to better optimised libraries than when the analysis is performed in reactant space. We have also shown that product-based analysis allows the simultaneous optimisation of multi-objectives in the design of libraries that are not only diverse but that also have 'drug-like' physicochemical properties.

0 0

Post a comment