Non-drug like molecules

Ajay Walters, W.P., Murcko, M.A. (1998) Can we Learn to Distinguish between Drug-like and Non-Drug-like Molecules Journal of Medicinal Chemistry, 41, 3314—3224. 

Ajay, Walters, W. P., Murcko, M. A., Can we learn to distinguish between drug-like and non-drug-like molecules ,/. Med. Chem. 1998, 43, 3314-3324. 

A comprehensive collection of 3,000 nontrivial drug-like molecules from the PDB is built by (a) excluding ubiquitous substrates, e.g., ATP, and (b) applying filters to exclude pockets with non-drug-like molecules (e.g., too large or too small) in the entire PDB Chemical Component Dictionary. That collection is merged with the above protein domain ensemble set to obtain multiple structure ensembles cocrystallized with at least one relevant compound. 

Of course, many more molecules in the CSD contain chemical fragments that are important in APIs and that provide valuable conformational or nonbonded interaction information despite being part of a non-drug-like molecule. In examining possible bioisosteres, the nonbonded interactions of common functional groups... 

Ajay, W. P. Walters, and M. A. Murcko,/. Med. Chem., 41, 3314 (1998). Can We Learn to Distinguish Between Drug-like and Non-drug-like Molecules ... 

Hann et al. used a similar, multilevel approach when pooling compounds for HTS. These authors applied three types of filters. Basic filters were used to remove molecules with non-drug-like features. Next functional group-based hard filters were utilized... 

As noted above, the field of combinatorial chemistry and multiple parallel synthesis started with libraries of peptides. In time, unusual residues crept into the products. While this evolutionis still ongoing, it is now accompanied by a major effort to produce libraries of small, drug-like molecules in library form. Many of the methods used for large molecules carry over but the largely non-iterative nature of small molecule synthesis is a significant complication. 

Recently the method employing DNP for recording heteronuclear 2D NMR spectra of small drug-like molecules was proposed by Ludwig and coworkers [101]. This method ensures significant improvement in sensitivity due to the high spin polarization, but limits the number of points sampled in indirectly detected dimension, so the combination with the non-uniform sampling scheme was necessary. 

The use of non-drug-like small-molecule substrates in the supersomal assay could be another factor contributing to the differences in DDI prediction. 

---