Cell decision making can be viewed as an information-exchange between cells and their microenvironment. We consider information as an organization principle for the spatiotemporal development of multicellular systems. Based on our MDFT approach (see below), we define a “free energy” functional that dictates the evolution of the system states in physical and phenotypic space, where the phenotypic part of the “free energy” involves the maximization of information theoretic quantities. Such an approach defines an integrative framework for studying the role of cell fate determination and phenotypic plasticity in development of multicellular systems. In particular, it allows for the quantification and analysis of information encoding, storage and transmission in a multicellular system [B5]. This theory allows us to identify the principles that dictate the cell decision-making [A33].
Biological lattice-gas ellular automata
Cellular automata (CA) can be viewed as simple models of self-organizing complex systems in which collective behavior can emerge out of an ensemble of many interacting "simple" components. They were introduced by J. von Neumann and S. Ulam in the 1950s in an attempt to model biological self-reproduction. I have mainly worked with lattice-gas cellular automat (LGCA), which is specific class of CA [A5, A8, A11]. In contrast to traditional CA, LGCA provide a straightforward and intuitive implementation of particle transport and interactions. Additionally, the structure of LGCA facilitates the mathematical analysis of their macroscopic behavior [B2, B4]. Recently, we have developed methodologies that allow for the derivation of physically- and data driven rules [A30, A34].
Multicellular Density Functional Theory (MDFT)
We are interested in a new mesoscopic framework based on an extended Dynamic Density Functional Theory (DDFT), for the first time, to the dynamics of living tissues by accounting for cell density correlations, different cell types, phenotypes and cell birth/death processes, in order to provide a biophysically consistent description of processes across the scales [A16]. The basis of this framework is the appropriate definition of a “free energy” functional. Recently, based on the DDFT model, we have developed a phase-field crystal (PFC) method for multicellular systems [A24].
Mechanical Multicellular Model (M3)
Mechanical interactions of cells play a crucial role in the cell decision-making during development of multicellular systems. Cell adhesion to other cells and to extracellular matrix (ECM) regulates many physiological and pathological events, such as migration, proliferation, differentiation and apoptosis. Moreover, mechanosensing can be pivotal to phenotypic plasticity of cells. Mechanical Multicellular Model (M3) describes intra- and inter-cellular forces, as well mechanical interaction with ECM in developing multicellular systems. We envisage the M3 allow us to understand the role of mechanics in phenotypic plasticity.
We regard the human body as a network of organs connected via the vascular and lymphatic systems. Cellular entities, such as immune cells, travel through this network and impact the functionality of each organ. Typical problems related to the organ network are metastasis or immune cell trafficking. To integrate the heterogeneous available data and predict the dynamics of the network of organs, we use a combination of machine learning methods and network theory.
Cell mitosis simulation with M3
Simulation of cell migation in ECM with M3
LGCA simulation of tumor spheroids
Pathway coupling effect on cell fate decision reliability
Many mathematical models rely on phenomenological relationships between model parameters and variables. While these relationships may be fitted to experimental data, they do not incorporate functions of directly measurable quantities at the cell scale. Multiscale models equip directly measurable quantities at the cell scale inform the model parameters at the continuum scale through upscaling techniques making multiscale models, in principle, predictive because the data used for calibration is distinct from that used for validation. Bridging these scales is a significant challenge.