Despite the high variability of structures from marketed drugs, molecules with biological activity have certain features in common. We have analysed these complex drug patterns to build up an expert system, which is able to discriminate between drugs and non-drugs:
MolScore-Drugs near 0: => lowest predicted probability
MolScore-Drugs near 1:
=> highest predicted probability
The expert system is based on a variety of reliable models. Structure-activity relationships allow the estimation of useful drug-like chemical space. Structure-property relationships which are derived from our in-house ADME/Tox-database are applied to predict ADMET properties and to identify potential risks in order to reduce clinical failures.
The results of MolScore-Drugs can easily be integrated into customer’s database. This will allow virtual screening, ranking and selecting compounds from external suppliers before they have to be purchased, see selecting compounds. MolScore-Drugs can be used to detect future blockbusters, see detecting blockbusters. All models which have been integrated into the expert system have been successfully validated with independent data sets.
MolScore-Drugs helps to identify and prioritise promising drug candidates with the maximum possibility of success in human trials, see lead prioritisation and selection.
All compounds with a valid chemical structure can be analysed; even before a compound has to be synthesised. The threshold (boundary between drugs and non-drugs) can be set by the customer. All results are supplied in an SDF-file. These results can be imported into other commercial software to actually see on screen (visualisation), for data analysis or for selection of compounds with additional filters.
Expert system to identify promising drug candidates
==> Lead selection
==> Prioritisation of drug candidates