Understanding the complex interplay between immune, infiltrating, and resident cells in rheumatoid arthritis (RA) is essential for predicting treatment responses and advancing personalised medicine. Our group has developed the RA Atlas, a comprehensive resource of large-scale molecular and cellular interaction maps (Singh et al., 2020; Zerrouk et al., 2022), which we are expanding to include diverse immune and tissue-resident cell types. Building on this foundation, we are constructing Boolean models to identify stable disease states, benchmarked against omics data and prior knowledge (Aghamiri et al., 2020). Recent advances include the development of one of the largest multicellular Boolean models to date, representing over 1,000 biomolecules across RA synovial macrophages, fibroblasts, and T cells (Zerrouk et al., 2024a; Zerrouk et al., 2024b).

To capture the spatial and dynamic aspects of RA, we are developing hybrid agent-based models (ABMs), where Boolean models guide cellular decision-making. This approach was first applied in the context of the COVID-19 Disease Map project (Ostaszewski et al., 2020, 2021; Niarakis et al., 2024). Our prototypes for RA, implemented in NetLogo and PhysiCell, simulate the joint microenvironment (unpublished work). ABMs offer outputs comparable to spatial transcriptomics, enabling validation against experimental datasets and providing a powerful tool to study the progression of key RA symptoms over time.

Our ongoing work integrates publicly available transcriptomic datasets to distinguish responders from non-responders to anti-TNF and JAK inhibitor therapies (Miagoux et al., 2021; He et al., 2025). By combining mechanistic modeling, multiscale simulation, and clinical data, this work lays the groundwork for a rheumatoid arthritis digital twin, capable of generating in silico predictions for experimental and clinical testing.

Jumeaux virtuels de patients pour une thérapie personnalisée par cellules CAR-T 

The novel field of engineered adoptive cellular immunotherapy (eACI) is quickly evolving. To improve outcome and availability of eACI, a variety of factors need to be to be addressed, including safety concerns, treatment resistance, limited manufacturing capabilities, and high treatment cost.

Virtual patient twins (VTs) enable a paradigm shift away from a one-size-fits-all application of eACIs towards a personalized, risk-adapted decision support. VTs for eACI manufacturing are in their early stages, but a whole-body VT to support decision making throughout the entire patient journey is lacking. As eACIs are living drugs, VTs in eACIs deviate from conventional biomedical VTs because they require to model the medicinal product as a dynamic biological system of living cells. Furthermore, the patient’s and the medicinal product’s path through eACI involve different stakeholders in distributed locations, thus the need for ensuring interoperability of VT components is high.

With the example of multiple myeloma patients eligible for CAR T cell treatment, I will outline and discuss the essential components for personalized virtual twin models in eACI (Weirauch et al. https://www.nature.com/articles/s41746-025-01809-6).

Good books