Risk Matrix Optimization
For broad safety in vivo endpoints, a compound may trigger any one or more of multiple mechanisms that lead to an adverse result. To predict these toxicities early in the optimization process accurately and inexpensively, multiple in vitro assays modeling the endpoint are utilized. Construction of a screening panel of predictive in vitro assays is a good strategy, however, this effort requires significant time and resources be invested. The identification of an optimal set of assays for accurate prediction can be a costly process involving testing of significant numbers of compounds and development of new assays.
CoRE™ makes this process more efficient by directing experimentation using active machine learning. With active machine learning, small subsets of compounds are selected cyclically to maximally increase the accuracy of the predictive models at the same time the panel is being optimized to predict the endpoint. As a result, only the most informative experiments are run ensuring that the most predictive assays are identified quickly. Historical data may be uploaded to CoRE for consideration as training data augmentation as well. In this case, CoRE will search the historical data as well as the KnowledgeBase for relevant predictive assays.
- Effectively guides assay and compound testing (avoids redundancy, saves time and resources)
- Automates data updating, model building and compound selection (saves FTE time)
- Easy to define and adjust panel profile as needed
- Panel prediction strength grows with use