300k diverse selection from ChemDiv 1.6M stock is based on Bemis-Murcko clustering workflow:
1) Generate Bemis-Murcko scaffolds (BMS) for every molecule in the entire 1.6M ChemDiv inventory
2) Apply REOS, MedChem & PAINS filters to remove reactive, toxic, promiscuous, and other undesirable structural motifs
3) Then apply Physico-Chemical Properties filters to remove non-druglike molecules
4) For the remaining selection, generate Bemis-Murcko scaffolds (BMS) for every molecule, calculate the number of unique BMS and number of molecules per each BMS
5) Split the selection into 3 categories for future clustering and diversity picking:
a. Number of individual molecules per unique BMS : 1-3
b. Number of individual molecules per unique BMS : 4-1000
c. Number of individual molecules per unique BMS : >1000
6) Iteratively select (i.e. cluster) individual molecules per each BMS within each category using RDKit MaxMin algorithm (Tanimoto, ECFP4, 2048 bits), cluster size (i.e. number of picked individual molecules per each BMS) calculated by the formula
7) Combine all 3 selection into the final 300k set
Medicinal and Computational Chemistry Dept., ChemDiv, Inc.
12760 High Bluff Dr, San Diego, CA, 92130
Phone: + 1 916 234 0888
Fax: +1 858 794 4931
You can buy this library or customize your own library from our discovery collection of 1 600 000 compounds and 60 000 building blocks.