Solubility library design

Several research groups have built models using theoretical desaiptors calculated only from the molecular structure. This approach has been proven to be particularly successful for the prediction of solubility without the need for descriptors of experimental data. Thus, it is also suitable for virtual data screening and library design. The descriptors include 2D (two-dimensional, or topological) descriptors, and 3D (three-dimensional, or geometric) descriptors, as well as electronic descriptors. 

Even though we have seen that the solubility predictions are poor for individual compounds, there is still a question over whether such predictions could nevertheless be useful in library design. We believe that they can. Figure 15.5 shows that QMPRPIus gets the general trend correct it shows that the proportion of soluble compounds increases with an increase in the solubility predicted by QMPRPIus (where a soluble compound is defined to have intrinsic solubility greater than 10 M). 

Physico-chemical properties and evaluation of potential safety liabilities are important aspects of the library design process. Predicted properties like hERG liability (45), compound aqueous solubility, etc. (46-48) have been extensively studied and included in various library design strategies (49, 50) as a part of multiple constraints optimisation. We have therefore further extended the ProSAR concept to take the library property profile into account in the design process. Several in-house calculated properties are considered these include a compound novelty check (that checks in in-house and external compound databases to see if the compound is novel), predicted aqueous solubility... 

Fig. 10.9. Aminoacetamide lead series used in establishing thresholds for solubility and metabolic stability to guide the library design. R1 can be adamantyl, cycloalkyl, benzyl or substituted benzyl, aryl, or heteroaryl. R2 can be alkyl, substituted alkyl, cycloalkyl, benzyl, substituted benzyl, or acetyl. R3 can be FI or OFF...

Next, we used an in-house library design software (see details in Chapter 15) to enumerate the virtual libraries and then calculated various physical properties. Products were removed from consideration if MW is > 300, number of rotatable bonds > 3, and ClogP > 3. For solubility, two in-house model calculations were applied as filters turbidimetric >10 mg/mL and thermodynamic solubility >100 xM. The resulting cherry-picked library was then reviewed by NMR spectroscopists to remove compounds with possible artifacts, likely to be insoluble, or likely to be false positive. These included some conjugated systems and compounds with likelihood of indistinct NMR spectra. 

Key words PGVL Hub, combinatorial chemistry, library design, reaction, synthesis protocol, reactant, product, enumeration, filtering, Chkl, kinase, inhibitor, SAR, ADME T (Adsorption, Distribution, Metabolism, Excretion, and Toxicity), selectivity, solubility, protein-ligand complex. 

Fig. 16.1. Progress of two rounds of Chk1 targeted libraries. Cpd-1 is the original HTS hit with a broad kinase inhibition profile and based on which the first round library was designed and synthesized. 1808-1 is the best hit from the first round targeted library with improved kinase selectivity profile, based on which the second round library was designed and synthesized. 1819-1 is the best lead with improved potency, kinase selectivity, and solubility. Co-crystal structures of Chkl kinase domain and corresponding lead compounds were solved and extensively utilized in structure-based singleton and library designs. For details of the X-ray co-Crystal structures, please refer to the publications from Ming and et al (4a) and Foloppe and et al for details (4b).

In the workflow used in this program (Fig. 1.3) primary screening is carried out in wafer format. The libraries are synthesized from soluble metal precursors using specialized library design software [8] and liquid-dispensing robots in a ter-... 

In summary, poor aqueous solubility is the single physicochemical property that is most likely to be problematical in a combinatorial library. It can be avoided in part by intelligent use of batch-mode solubility calculations. The solubility problem is not simply a technical issue in library design. It is exacerbated by chemistry synthesis considerations and by the timing of the availabihty of combinatorial exemplars. Formulation fixes are available unless the solubility is extremely poor, but these should be avoided as much as possible. Poor permeability is seldom a problem in combinatorial libraries, but is disastrous if present since effectively formulation fixes do not currently exist. 

---