About his PhD project

Progress in microscopy technology led to a burst in the generation of data on the dynamics of different biological systems, from proteins to entire microorganisms. However, this data is always imperfect due to multiple experimental limitations, which make the inference of underlying models a formidable task. It is in our best interest to make up for these data collection limitations by optimizing the inference methods used to process this data. For instance, some advances have significantly improved inference from data corrupted by measurement noise [1]. In this project, we will tackle the problem of incompleteness of experimental data. Considering complex biological systems, it is never possible to measure every degree of freedom that is relevant to the observed dynamics. Therefore, when proposing a model for the observed dynamics, one needs to be aware that the dynamics of the measured quantities also depend on unmeasured hidden variables. Still, current inference methods often ignore the existence of these hidden variables. We will further develop state-of-the-art inference methods to consider unmeasured hidden variables that may affect the dynamics of observed quantities. Using information theory, we will develop tools to estimate the amount of information about hidden variables available through the measured ones. Given enough information in the available data, our improvements will permit researchers to uncover the hidden structure of biological processes, improving their understanding. In the absence of enough information to reveal this underlying structure, at the very least, we will manage to estimate temporal correlations in the observed quantities induced by the dynamics of the hidden variables, allowing improved predictions for these systems.

[1] Frishman, A. & Ronceray, P. Learning Force Fields from Stochastic Trajectories. Phys. Rev. X 10, 021009 (2020).