In this webinar, Matt Segall and Samar Mahmoud describe the generation and validation of a ‘global’ deep learning model for drug discovery using imputation on a data set of 300,000 compounds and 500 experimental endpoints, targeting global health indications.
We demonstrated how this deep learning global model can be applied to individual optimisation projects, offering improved compounds design performance over ‘local’ project-specific models by learning across a broad chemical diversity
Even including unrelated endpoints from many projects does not cause a loss of performance for individual projects and endpoints – an example of “build once, run everywhere.”
The webinar will provide:
- An introduction to deep learning imputation using AlchemiteTM
- An overview of data set and objectives
- A summary of model validation
- An example of global deep learning model to project optimisation
