A An optimized BEI SEIscore based on
(A) An “optimized BEI-SEIscore” based on the trends, emerged from the statistical analysis of several optimization studies in pharmaceutical industry (Abad-Zapatero, 2007, Abad-Zapatero and Blasi, 2011). This type of optimized scores has been used in many previous studies.
Ki=1nM at 300° K, compare to binding vx 765 of −12.4kcal/mol.
N (non-hydrogen atoms)=25, with average molecular weight of 13.32Da
Vander Waal PSA=50Å2
BEI=26 and SEI=18.5
This 26/18.5 (BEI/SEI) score can be considered as drug likeness score of a reference drug. Since drug optimization is a multi-step process and generally BEI-SEI score improves with each step from hit to lead and then to drug, some relaxation (of about±4 and±3.5 respectively for BEI and SEI) should be given to our standard score of 26/18.5 for initial hits (Table SS6, ESI). Hence molecule with BEI score of 22–30 and SEI score of 15–22 can be considered as a good starting point. Drug-likeness of all other molecule depends on their location and proximity to reference 26/18.5 score in efficiency plane. Efficiency indices of all the known inhibitors (PYR, CYC, WR99210, TMP, MTX, INH1 and INH2) and substrate DHF were also calculated and found well within (except MTX) the above mentioned cut off with average BEI and SEI score of 25.8 and 15.8 respectively (Hasting and Sibley, 2002, Sirichaiwat et al., 2004, Toyoda et al., 1997).
(B) Hits with contact footprints similar to the known inhibitors or true active compounds have greater chances to turn into competitive inhibitors. So in pose-clustering, interaction profiles of docking poses were compared with six well known antifolates (PYR, CYC, WR99210, TMP, MTX, INH1 and INH2) and substrate DHF. Finally all the compounds were grouped together on the basis of their contact footprints (interaction profiles). Compounds presented in high priority cluster (cluster with maximum no. of known actives/inhibitors) should be given priority over those present in other clusters.
Conclusion The present study demonstrated the power of computational tools to access the behavior of compounds in optimization plane. Many trends were established on the basis of efficiency indices and pose clustering. These investigations are particularly useful for libraries emerging from highly efficient and diversity enabling reactions such as MCRs. Authors believe that prior computational exploration like this before wet lab and actual biochemical assay would help medicinal chemists in managing time, resources, chemicals and in prioritizing their work efficiently.
Acknowledgements One of the authors MK would like to thank Council of Scientific and Industrial Research (India) for SRF fellowship. This work was financially supported by CSIR, New Delhi [02(0123)/13/EMR-II dated12.4.2013].