Physics of Biological Data Analysis

Organizers:
*Marianne Bauer, Delft University of Technology
Anne-Florence Bitbol, École Polytechnique Fédérale de Lausanne
Ilya Nemenman, Emory University
Greg Stephens, Vrije Universiteit Amsterdam

The amount and quality of biological datasets available is fast increasing. This holds at the molecular and cellular scales (gene sequence and expression data), but also at larger ones (populations and communities of cells, behavior of one or multiple animals). Making sense of these extraordinarily rich datasets calls for new inference and interpretation methods, and ultimately new theory. Since microscopic theories of biology may not provide insight on larger scales, data-driven approaches play an integral role in biophysics, for example guiding us to discovering new physical laws.

Various analysis approaches inspired by statistical physics and by machine learning are currently being developed. They range from fitting models to model-free analysis, and include supervised and unsupervised approaches. Recent advances in machine learning offer powerful new methods. For instance, some deep neural networks capture very well the rich structure of biological sequence data. Physics-based concepts play important parts in these analysis approaches, including deep learning ones. In turn, these models provide insight on biophysical phenomena.

This workshop will bring together scientists modeling biological data, performing and analyzing data-rich experiments, and those who are interested in developing, using and understanding new data analysis methods, such as deep learning. We will compare approaches, discuss successes and failures in data analysis, and reflect on future directions.