Pour suivre la conférence à distance, cliquez ici.
Abstract:
The concept of patient trajectory is important in understanding a disease as a dynamic process that evolves over time. Evaluating patient trajectories typically involves longitudinal (diachronic) datasets, which are still relatively scarce and expensive. Drawing from the concepts of pseudotime and the ergodic principle in single-cell data science, we recently introduced a methodology for quantifying hypothetical stereotypical patient trajectories from synchronic (snapshot) clinical datasets. These trajectories characterize the landscape of all possible disease conditions, illustrating how a specific disease state develops along stereotypical routes, marked by “points of no return” and “final states” such as lethal or recovery states. We demonstrate how applying this methodology yields insights from extensive clinical datasets in cardiology and COPD, which can be used to enhance these diseases management.
Short bio:
Andrei Zinovyev has a background in computational biology and machine learning, focusing on developing new methods and algorithms for modeling and analyzing biological data. After completing a 3-year postdoc at the Institute des Hautes Études Scientifiques (the French counterpart of the Advanced Studies), he transitioned to the Institut Curie, where he co-founded the “Computational Systems Biology of Cancer” research group, which he co-led for 18 years. In 2019, Andrei got a chair at the Paris Artificial Intelligence Research Institute, focusing on the application of machine learning to multiomics and clinical data. In 2022, he made the switch to industrial research in the role of Principal Scientist at Evotec in Toulouse in the In silico R&D department.
