Machine Learning framework and models for predicting clinical trial quality and drug response

Offentlig prosjektsammendrag

Currently most used drugs in oncology are ineffective for a large proportion of patients according to U.S Food and Drug Administration. As a response to this problem PubGene will in this project develop a generic machine learning framework and two key machine learning models, and verify if the framework and models can be used in future product developments within precision medicine decision support. The focus are models for predicting individual and/or population-based drug response based on patient profiles and predicting the quality of a clinical trial and what parameters are most important to predict if a clinical trial will result in a drug or procedure approval. The clinical trial model can be used by hospitals when stratifying which clinical trial a patient should be included in if there are multiple alternatives, while pharmaceutical enterprises can use the model designing better clinical trials increasing the probability of achieving approval. The drug response prediction model can be used by pharmaceutical enterprises to select clinical trials population, while the hospitals and clinics could use the model to better select which drugs and which doses to use for each individual patient.