SDSC-AI4Science seminar: Modeling the collective dynamics of in-vitro populations of neurons

We are happy to announce the SDSC - AI4Science monthly seminar, a seminar co-organized by the EPFL AI4Science Initiative and the Swiss Data Science Center and focussing on projects in which data science, statistics, machine learning and AI are applied to the sciences. Each seminar will feature a presentation of one applied project, geared towards an audience with expertise in Data Science methods, from the initial formulation of a research question in science associated with sources of data, to the model, algorithms and analyses produced. The presentation will be highlighting the choices made, the challenges encountered, interesting technical questions and possible further developments. A number of the projects presented will be collaborative projects of the Swiss Data Science Center. One of the objectives of the seminar is to foster exchanges between researchers working in methods and applied data science research in the sciences, and to create new opportunities of collaborations.

Each session will feature a talk followed by a discussion with questions from the audience.

The first seminar will take place on November 16th at 16h00 in CM 1 3.

Speaker: Christian Donner, Senior Data Scientist at SDSC

Title: Modeling the collective dynamics of in-vitro populations of neurons

Abstract: In the DeepEphys project we aim at improving our understanding of the neurophysiological basis of Parkinson’s disease (PD). To this end, we recorded in-vitro activity from healthy and PD neuronal cultures that have been derived from induced pluripotent stem cells. The talk’s focus will be on how we can use such large-scale recordings to characterise the collective neuronal behaviour. I will present two approaches that we followed: First, using discriminative classification models we identify distinctive physiologically meaningful predictors for PD. In the second part, switching gears, we focus on learning a generative process of the observed data. More precisely, we will consider a Cox process model for the observed neuronal activity, i.e. a doubly stochastic point process, where the neurons’ firing rate is modelled nonparametrically by a Gaussian process. The Gaussian process is defined over a space of the past neuronal activity. Once the model is learned via variational inference, we can derive a set of differential equations for the neuronal population, which allows us to use tools from dynamical systems analysis to gain insights into the dynamics of the underlying neuronal culture. This way we can formulate experimentally testable hypotheses for the future.