Vendredi 18 octobre 2019 15h00 – 17h00
Salle TD 9.02, Bâtiment 9 – Université de Montpellier, Campus Triolet
The development of high throughput sequencing techniques currently generates an increasing volume of data at moderate costs. These data are often of very high dimension, heterogenous and simultaneously acquired at different levels of a living organism. A challenging question is how to integrate all these data , e.g., how to combine these different views of the living functionning and how to combine them with external information to better understand the underlying biological mecanisms at stake or to better predict a quantity of interest (a phenotype for instance). Among the methods developed to address this issue, kernel methods have several appealing properties that make them a popular approach. In this talk, I will present the general setting of kernel methods and how they can and have been used in computational biology. I will more specifically focus on exploratory methods (or unsupervised methods) for data integration and will illustrate my talk with an example taken from TARA Oceans datasets.
Parts of this talk are related to works co-authors with Madalina Olteanu, Fabrice Rossi and Jérôme Mariette.