Diversity, combinatorial libraries descriptors

The descriptor uses readily calculable physicochemical properties from the topological structure. The descriptors used in this study were atomic weight, hydropho-bicity, molecular refractivity, atomic charge, polar surface area, hydrogen bond acceptors, and hydrogen bond donors. The authors note that Martin et al. [32] applied a similar approach for the design of diverse combinatorial libraries. 

While not exactly the same as the methods described above in that DOE cannot be applied retrospectively to diverse datasets, it has been used very successfully to guide the selection and evaluation of compounds from combinatorial libraries (59,60). However, DOE has been successfully applied only in cases where limited libraries of related compounds (e.g., peptides) were being evaluated. The reason for this is intuitively obvious, as one of the assumptions of DOE is that variability in the descriptors is continuous and related to activity over a smooth response surface, so that trends and patterns can be readily identified. With HTS data both of these assumptions are generally not true, as molecules can display discontinuous responses to changing features, and the SAR of even related compounds does not map to a smooth continuous response surface (for example, Fig. 2). 

The ten molecular descriptor PCA analysis of a subset of three libraries performed by Feher and Schmidt18 suggested that the combinatorial library clustered very tightly and represented significantly lower diversity than natural products or drugs. The drugs and natural products showed wider diversity and covered a diversity map that was not represented by the combinatorial library. 

An additional consideration on the choice of descriptor for diversity analyses is the speed with which the descriptor can be calculated since diversity studies are often applied to the huge numbers of compounds (potentially millions) that characterise combinatorial libraries. Thus, some computationally expensive descriptors such as field-based descriptors [28] or descriptors derived from quantum mechanics [29] are not appropriate for diversity studies. 

Diversity The unrelatcdncss" of a set of. for example, building blocks or numbers of a combinatorial library. Measured using physicochemical or structural descriptors, a. set with high diversity. spans a larger fraction of "chemical space." Cluster analysis is one technique used to quantify diversity. 

The descriptors most commonly applied to measure diversity in large compounds sets (combinatorial libraries or screening collections) include 2D fingerprints (for sub-structural diversity) and 3D three-point pharmacophores (to measure the range of pharmacophoric groups that are present within the compound set). 

Most important for a successful diversity description of a combinatorial library is the appropriate choice of the descriptor set Thus far, a large number of different descriptor sets are pubhshed or implemented in software packages sold commer-... 

Fig. 13 Selection of compounds from a virtual combinatorial library, a) First six steps of a maximum dissimilarity selection, b) Selection by excluding similar compounds, c) Maximum diversity selection as a result of clustering, d) Grouping by partitioning of descriptor space.

Steric fields derived from the CoMFA method [75] were used as 3D shape descriptors, while their major drawback is that a reliable superposition rule for databases is required. Thus, topomeric fields as significant enhancement for comparing the diversity of a set of variable groups attached to a common scaffold have been developed and validated [76], Those fields rely on canonical conformations of aligned compounds from combinatorial libraries using the reaction core. They use a novel way to encode flexibility depending on the shortest bond distance between any atom to the reaction core. 

In an effort to compare descriptor distributions between compounds from different sources and synthetic paradigms, Feher and Schmidt [121] used PCA-based methods to compare property distributions from natural products, drugs, and combinatorial libraries. In this case, the authors used chemical space as a common framework to ask questions about the how the origins of compounds are manifest in their structural features at a global level. In particular, this study demonstrates the general dominance of synthetic efficiency, rather than structural diversity, in the preparation of compounds by combinatorial chemistry. The descriptors most able to distinguish natural products from those synthetic molecules studied were those that rendered the latter class easier to make, such as fewer... 

---