The data available in drug discovery typically have a high degree of uncertainty, due to experimental variability or predictive error, making it difficult to decide with confidence which compounds to prioritise or whether it is appropriate to 'kill' a chemical series or project.
Within StarDrop, the project team can define a scoring profile, specifying the success criteria and their relative importance for any property data that is available, whether predicted or experimental. In addition to simple pass/fail thresholds, ranges and trends may be defined in order to more subtly reflect the acceptable trade-offs between properties.
StarDrop's unique probabilistic scoring algorithm assesses all of the compounds against this profile, taking into account all of the available data and, crucially, the uncertainty in each data point. The result is a score for each compound, representing its likelihood of success against the required criteria and an uncertainty in the overall score allowing confident decisions to be made when the data distinguish compounds with confidence.
