成人VR视频

Integrating large- to high- dimension markers in mechanistic models

Melanie Prague,聽Universit茅 de Bordeaux
Tuesday April 14, 12-1pm
Zoom Link:听
In Person: 550 Sherbrooke, Room 189

Abstract:听Mechanistic models are widely used to describe and explain biological processes over time. However, they typically rely on a limited number of observable compartments and sparse longitudinal data. As a result, these models are often either too simple to capture complex biological phenomena or they face identifiability issues, particularly when considering interindividual variability in the form of nonlinear mixed-effects models based on systems of differential equations. In parallel, with ongoing technological advances, longitudinal high-throughput data (e.g., -omics, including transcriptomics and proteomics data) are increasingly available in various contexts and could bring valuable information into mechanistic models to better capture underlying biological processes. However, integrating such high-dimensional data to inform the dynamics remains a major challenge, both mathematically and for broader interpretation in public health applications.

In this talk, I will present two complementary approaches for integrating large- to high-dimensional biomarkers into mechanistic models. The first approach, called lasso-SAMBA, addresses robust covariate selection in ODE-based non-linear mixed-effect models. It extends the original SAMBA algorithm (which is an iterative model-building algorithm that fastly and sequentially identifies relevant covariates on parameters while estimating model using the SAEM algorithm) by replacing stepwise inclusion with Lasso regression combined with stability selection, ensuring a more reliable identification of relevant covariates while preserving the monotonic decrease of the information criterion. The second approach uses observed -omics data to infer the dynamics of unobserved immune compartments. It relies on an iterative algorithm that alternates between a regularization step, which identifies the most informative biomarkers through penalized likelihood derivatives, and a mechanistic inference step, where population parameters are estimated using the SAEM algorithm in Monolix. This framework enables the selection of biomarkers whose temporal patterns best reflect the latent compartments in our model.

Together, these methods provide powerful tools for integrating and selecting high-dimensional biological data in mechanistic modeling. They will be illustrated on exemples of immune dynamics after vaccination for Varicella-Zoster virus and immune-viral dynamics after SARS-CoV-2 infection.