Lucie Brolon
Modeling T-cell and dendritic cell communication
When an infectious agent enters the body, it is detected by dendritic cells (DCs), which emit biochemical signals to initiate an immune response. These signals induce the differentiation of T cells (TCs), which subsequently release signals to combat the infectious agent.
Modeling the relationship between the signals emitted by DCs (input) and those of TCs (output) is challenging. DCs can produce diverse signals, multiple DCs may emit identical or distinct signals simultaneously, and TCs process combinations of these signals. A single DC signal can have varying effects on TC signals depending on the simultaneous signals emitted by other DCs—referred to as the context. Each signal must therefore be analyzed within its context.
To address this challenge, we developed a hybrid method that combines two algorithms: the regression tree PILOT [1] and the random forest SIRUS [2]. In this talk, I will introduce this new model.
[1] J. Raymaekers et al. “Fast linear model trees by PILOT”. In:Mach Learn113 (2024),6561–6610. [2] C. Benard et al. “SIRUS: Stable and Interpretable RUle Set for classification”. In:Elec-tron. J. Statist.15(1) (2021), pp. 427–505.
