[Statlist] Research Seminar in Statistics | *FRIDAY 31 MARCH 2023* | GSEM, University of Geneva

gsem-support-instituts g@em-@upport-|n@t|tut@ @end|ng |rom un|ge@ch
Mon Mar 27 09:30:57 CEST 2023


Dear all,

We are pleased to invite you to our next Research Seminar, organized by Professor Sebastian Engelke on behalf of the Research Center for Statistics. (https://www.unige.ch/gsem/en/research/institutes/rcs/team/)

FRIDAY 31 MARCH 2023 at 11:15 am, Uni Mail M 3393 & Online
Zoom research webinar: https://unige.zoom.us/j/92924332087?pwd=U1U1NFk4dTFCRHBMeWYrSDBQcXBiQT09
Meeting ID: 929 2433 2087
Passcode: 399192

Fast Bayesian Joint Inference for Uncovering Disease Mechanisms at Scale
Hélène RUFFIEUX, University of Cambridge, UK
https://www.mrc-bsu.cam.ac.uk/people/in-alphabetical-order/n-to-s/helene-ruffieux/

ABSTRACT:
The landscape of biomedical research is fast evolving, as new molecular data paradigms emerge and provide greater granularity for understanding disease mechanisms. We are also increasingly aware that pathogenic responses are tightly coordinated at an organismal level, which calls for modelling approaches that can provide a holistic understanding of complex interplays across biological systems.

Bayesian hierarchical modelling offers a principled framework to leverage complicated dependences within and between heterogeneous biological data sources, while conveying uncertainty coherently. However, the tension between flexible joint modelling and practical inference at the scale required by current analysis needs remains substantial, hence modelling and algorithmic aspects need to be considered jointly.

Motivated by research questions on infectious and immune-mediated diseases, I will present three examples of hierarchical modelling approaches aimed at clarifying disease mechanisms at scale. These approaches rely on models that lend themselves to efficient inference and implement inference algorithms that are model specific. They cover sparse regression modelling, graphical modelling and longitudinal latent modelling, with as common threads (i) the borrowing of strength across multiple conditions (molecular entities, tissues, cell types or disease subtypes) through the model hierarchy, and (ii) the use of deterministic inference procedures with expectation-maximisation and structured variational schemes that improve exploration of multimodal posterior spaces. All three approaches use context-tailored prior formulations, such as the horseshoe prior which shrinks noise globally, hence accommodating the highly sparse nature of molecular effects, while leaving individual signals unshrunk thanks to heavy-tailed local-scale distributions.

> View the Research Seminar agenda: https://www.unige.ch/gsem/en/research/seminars/rcs/

Regards,


Marie-Madeleine

Marie-Madeleine Novo
Assistant to the Research Institutes
gsem-support-instituts using unige.ch



More information about the Statlist mailing list