Collecting chemical structures from providers

The first step of the creation of a virtual chemical database is to collect structure files from a maximum of providers. There is a lot of providers and they often give structure files freely on the web. Here is some providers :

Remove non-unique structures

Once the collected files are inserted in the software, the non-unique structures must be removed. These structures represent an important part of the database, and they slow down virtual screenings and HTS. The software use the IUPAC Chemical Identifier in order to remove the doubloons. This code is in a beta version but it is very usefull. It give an unique code to a molecule, like CAS number, but contrary to CAS number it does not have to connect to a database to give the unique code. The code is compute from the structure file of the compound. It can handle Z/E an R/S stereochemistry, isotopes and tautomeric forms.


Drug-like properties

Several descriptors are computed when molecules are inserted in the software : number of H donors, number of H acceptors, logP, MW, number of rotatable bonds, number of halogens, number of O and N, number of SSSR (Smallest Set of Smalest Ring), size of the biggest ring, number of single chains, number of perfluorinated chains, number of reactive functions, presence of warheads and promiscuous aggregating inhibitors.


Structure corrections

The structure given by the providers must be corrected. The H must be added. The counter-ions must be deleted (it can disturb docking tests). And last, the structures must be minimized. ScreeningAssistant can use Corina for structure minimization.


Technical details

The project is programmed in Java. The datas are stored in a MySQL database. Chemical structures are handled by a free Java API : JOELib. The program uses this API to parse SDFiles and compute descriptors. Optionally Marvin can be used as molecular viewer. JFreeCharts allows to display statistics charts of the database. JGoodies is used for the graphical interface.