When it comes to virtual screening , an important decision to make is which compounds to include in the study. The ZINC database has emerged as an immensely valuable tool for prioritizing and downloading specific compounds in computationally convenient forms. I am very impressed with the excellent organization, clear instruction, and overall reliability of the database. A minor complaint I have is that over time the claim that a substantial portion of the database is “available for purchase” within a month or so seems… hard to believe, especially as my retrosynthetic intuition has progressed. Also, the compounds which were deposited more than 10 years ago may not be stable enough to last more than 10 years without some minor decomposition that could interfere with biological testing. The newer compounds in the database are naturally more and more complex than their predecessors. The increase in complexity makes synthesis more difficult and thus renders the compound less likely to be purchasable as well. In any case, I recently used the database for virtual screening of allosteric inhibitors of Leishmanial methionyl-tRNA synthetase as well as competitive inhibitors of Epidermal Fatty Acid Binding Protein (FABP5). The experiment with FABP5 was a great example of decision making within the tranches, or molecular space designations, of the ZINC database.
The tranches of the ZINC database are organized primarily through molecular weight and lipophilicity (via calculated Log P). This provides a distinct section chemical character to the types of molecules one can choose for docking. Additional descriptors include filters based on presence of reactive functional groups, charge of ligands, and ease of purchase. As you might imagine, knowing your target naturally leads one to prioritize certain chemical space for docking and virtual screening. This leads me to the natural question of what kind of chemical space I should explore for my virtual screening of a FABP inhibitor. For discovery of a Fatty Acid Binding Protein inhibitor, I split my virtual screen libraries into two distinct tranches. Both tranches include only molecules with at least 1 negative charge, however I chose one library to be a smaller molecular weight, more polar compounds. The second library was higher molecular weight, and more non-polar. My rationale for splitting the chemical space this way is as follows; The heavier, more non-polar library should yield true “derivatives” of fatty acids with very similar binding profiles as the cognate ligands, whereas the smaller, more polar library will demonstrate a more unique binding profile to the cognate ligand. In both situations, the negatively charged moiety should be interacting with ARG 106, ARG129, and TYR 131. I presume that there will be greater selectivity for compounds from the lighter, more polar tranche due to more precise hydrogen bonding with one FABP isoform over the other. I will be sharing a subsection of the results in an upcoming post. Until next time, enjoy.
