This work has been realised in a close collaboration of van der Schaar lab, CCAIM and AstraZeneca.
Clinical pharmacologists understand that patients are unique. Every patient has a biological system that responds differently to drugs and adjusting dosing to specific circumstances is often necessary. In clinical trials, trial and error can be time-consuming, and waiting weeks to learn how a patient responds to an investigational medicine is not always feasible. Mathematical models may be able to help make predictions about patient outcomes. However, what model should be used for the situation at hand?
Synthetic Model Combination (SMC), a machine learning method for constructing new model ensembles, is our newest answer to that question. SMC builds a new ensemble weighting existing models according to their likelihood to accurately represent a novel case. Based on our results, SMC is more robust and gives more accurate predictions than existing models, especially when there is no new data to judge which existing model is best. SMC can be reverse engineered to the needs of the situation and find applicable data for individuals who do not fit any of the existing models. This has the potential to really benefit – under-represented groups and special cases such as rare diseases and unique biological responses.
Pulling from a wide array of available models, when compared to trial-and-error and other methods: “SMC simply gives the best chance to get it right the first time”, Prof Richard Peck (CCAIM/University of Liverpool) says.
For drug developers, this is an exciting application as SMC provides a potential method to maximise the value of existing and new models to make more accurate predictions for new populations and new patients, ultimately improving the understanding of drug effects and how patients should be treated efficiently.
Prof Mihaela van der Schaar (CCAIM/University of Cambridge) has the following to say about the potential impact of SMC:
“I am tremendously excited about the possibilities that arise when machine learners and pharmacologists work together. By combining the power of cutting-edge technology with deep insights into the workings of the human body, we can unlock a new era of precision medicine. With more accurate predictions, clinicians will be able to develop new drugs more efficiently, and provide more effective personalised treatments, ultimately improving outcomes for patients around the world.“
SMC is flexible and can be used to understand and address a wide range of biological systems, pharmacological questions, and prediction-based approaches. As with all ensembling methods, SMC is only as good as the models fed into it but for now, SMC outperforms other ensembling methods. A big leap forward for drug developers and patients alike.
Comment from Megan Gibbs, (Vice President and Global Head, Clinical Pharmacology & Quantitative Pharmacology, AstraZeneca) :
“This publication is just one example of the potential for integrating AI into clinical pharmacology. Through closer collaboration between data scientists and clinical pharmacologists, we are able to improve our understanding of disease and its direct impact on patients’ lives.”
For a deeper understanding of how SMC works and how it can impact pharmacology, watch this short conversation between Prof Richard Peck and Alex Chan, leading contributors to the article:
You can find the full paper here.