Drug space

Compounds from traditional drug space ( common drugs —readily available from chemical suppliers), often chosen for studies by academic laboratories for assay validation and computational model-building purposes, can lead to misleading conclusions when the results of such models are applied to real discovery compounds, which most often have extremely low solubilities [16]. 

While this may in fact be the case for natural product mixtures, it is rarely the case when dealing with synthesized mixtures. Despite our attempts to create real molecular diversity in the test tube, our efforts have not even begun to anticipate the true diversity of atomic connectivity within "drug space" (estimated to be of the order of 1063 unique compounds, theory, famously in this case, greatly outpacing the amount of matter in the universe). Thus, combinatorial chemistry was never practically able to produce true chemical diversity and compounds produced in such library format ended up looking very much like one another, with the attendant similarities in biological activity profiles. 

So far all the models introduced above utilized relatively smaller data sets. One potential problem of these models is that the model applicability may be limited, since only a tiny fraction of drug space is represented by a small set of molecules in the training set. Recently, more and more models have been developed using much larger data sets, such as Delaney s model (2,874 compounds) [34], Votano and Parham s model (4,115 compounds) [35], Obrezanova et al. s model (3,313 compounds) [36], and Wang, Hou, and Xu s model (3,664 compounds) [25]. 

Not all lead spaces contain areas of activity that qualify as drug spaces. Finding a molecule within one of the drug spaces is the goal of a drug discovery program.13... 

FIGURE 9.19 Walking a molecule from hit space to drug space by six sequential structural changes... 

A new approach termed Biology Oriented Synthesis (BIOS) has been developed recently by Waldmann et alf This approach is based on the structural similarity between small bioactive molecules on the one side and their receptors, that is proteins, on the other side as well as on the complementarity of both. BIOS employs compound classes from biologically relevant regions of chemical space, for example natural product or drug space, to select scaffolds as starting points for the design and synthesis of small focused libraries with limited diversity. In this respect BIOS provides a conceptual alternative to other approaches... 

Drug space map coordinates are the t-scores extracted via Principal Component Analysis. PCA was performed on a total set of423 satellite and core structures described by 72 descriptors representing size, lipophilicity, polarizability, charge, flexibility, rigidity, and hydrogen bond capacity. 

Selected molecules include a set of satellite structures and a set of representative drugs ( core structures). Satellites, intentionally placed outside drug space, have extreme values in one or several of the desired properties, while containing drug-like chemical fragments. 

Numerous chemical filters were apphed to focus the training space to relevant drug space. Therefore, reactive fragments, such as aliphatic sulfates, as well as rarely populated chemical fragments were excluded. 

Bade, R., Chan, H.R, Reynisson, J. (2010) Characteristics of known drug space. Natural products, their derivatives and synthetic drugs. Eur. J. Med. Chem., 45(12), 5646-5652. 

---