Often scientists may not know the exact biological mechanisms dictating the data signatures. For example, which risk factors drive responders to a cancer treatment or how a biomarker is related to clinical outcomes. Current data science methods require big data, and they ignore prior knowledge of the problem at hand. PumasAI is poised to disrupt this. DeepPumas enables seamless integration of domain-specific knowledge and data-science methodology, reducing dependence on data size and enabling faster decision-making.
Here, we will learn, hands-on, how DeepPumas can automatically discover complex predictive factors to individualize predictions. Furthermore, we will learn how dynamical systems that model the longitudinal evolution of patient outcomes can be augmented by machine learning – enabling data-driven discovery of the underlying biology. Together, this enables effective use of data to rapidly develop models that predict individual outcomes from heterogeneous sources of patient data.
Applicable across the whole chain of drug development, from lead generation, quality by design manufacturing, clinical research, and market research to individualized patient management, DeepPumas is not an incremental improvement but a game-changer.
During the first day we will learn
- How to use Pumas for pharmacometric modeling
- Basics of machine learning
- Neural ODEs and universal differential equations
During the second day we will learn
- How to leverage random effects to deal with longitudinal data
- How NLME models are closely related to generative AI
- How to leverage machine learning in NLME models - DeepNLME
- How to use DeepPumas for data-driven discovery of predictive factors
- Niklas Korsbo - [email protected]
- Mohamed Tarek - [email protected]
- Andreu Vall
- Jose Storopoli
This content is licensed under Creative Commons Attribution-ShareAlike 4.0 International.
