Combining Machine Learning Technology with a Complex R&D Challenge to Invent a New Way of Discovering Medicines
An R&D Leadership team wanted to explore how might machine learning accelerate the identification of novel
We designed a multi-day session that was divided into 3 parts:
To prime the participants, we incorporated: a data scientist panel, embedded machine learning and AI Subject Matter Experts into teams and integrated outside case studies to spark fresh solutions.
A 10 week experiment was designed with IBM Watson to extract scientific evidence and non-obvious linkages for promising treatment combinations.
Pfizer/IBM co-developed models that identified a rank ordered list of promising combinations for further assessment and demonstrated the potential to identify promising new candidates from the literature, sometimes years prior to definitive publication. The approach potentially provides PFE Researchers with a head start.