Compound Libraries Available for High-Throughput Screening (HTS) in the Stanford HTBC

Our compound library contains over 130,000 diverse compounds from ChemDiv SPECS & BioSPECS, Chembridge, ChemRX, LOPAC, NIHCC, and NCI DTP. Since our emphasis is on diversity, most of our compounds are not from combinatorial nor directed libraries. We screen one compound per well in a 384 well microplate format (~400 384 well plates). List of Available Screening Compounds

The initial 30,000 compounds from Specs were selected on the following criteria, with the help of Anang Shelat of R. Kip Guy's lab fomerly at UCSF. Anang used InHouse models and the commercially available, Scitegic Pipeline Pilot, to perform the computational analysis. The 50,000 compounds from ChemDiv were selected similarily by Brian Wolff in collaboration with Jan Williams, formerly at the Small Molecule Discovery Center-HTS Division, and James Wells at UCSF. Additionally, we purchased a 10,000 compound Kinase-directed library from ChemDiv with funds from a generous donation by a private foundation. For additional information on screening compounds see PubChem.

(A) Using SD files (structure files containg over 200,000 compounds available) from Specs, molecules were passed through a standardization procedure: charges were cleared and set to formal charge, salts were stripped, certain topologies (such as nitro, sulfate) were canonicalized, and a canonical tautomer was selected.

(B) These molecules were passed through a Lipinski "Rule of Five" filter: Num_Atoms > 0 AND (N_count and O_count <= 10) AND (100 <= MW >= 500) AND (Num_H_Donors <= 5) AND (-5 <= AlogP <= 5) AND All Organic Atoms (salts were stripped earlier, so they are not considered here). This resulted in about 150,000 molecules.

(C) These modified Lipinski molecules were ionized using an InHouse pka model and filtered based on formal charge: -3 <= FC <= 3. Nearly all the molecules passed.

(D) The molecules were then passed through a REOS (Rapid Elimination of Swill) filter (eliminating functional groups deemed reactive by literature and consultations with medicinal chemists). This step eliminated another 30,000 molecules.

(E) A Bayesian categorizer was used to distinguish the Specs molecules from UCSF's InHouse Library (input variables: chemical fingerprint and calculable physical properties). The compounds from Specs were annotated with this score (lower score = most unlike InHouse compounds), thus creating a diversity spread of the library.

(F) A randomized selection of the ~120,000 compounds was ordered to comprise the inital 30,000 compounds of the HTBC's library.

References:

Christopher A. Lipinski, Franco Lombardo, Beryl W. Dominy, Paul J. Feeney "Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings", Adv. Drug Delivery Rev., 1997, 23(1-3), 3-25.

Brian Y. Feng, Anang Shelat, Thompson N. Doman, R. Kip Guy, & Brian K. Shoichet "High-throughput assays for promiscuous inhibitors", Nature Chemical Biology, 2005 Aug;1(3):146-8.