About Daphne

Initially, development was focused on the first-generation software platform (PlazaSur). Plaza Sur provided a working model of germinal center dynamics that included important B-cell dynamics, such as:

  • proliferation
  • somatic mutation
  • differentiation between centrocyte and centroblast phenotypes
  • migration between light and dark zones in response to chemotactic gradients
  • interaction with Follicular Dendritic Cells (FDC) in the light-zone (including B-cell Receptor (BCR) affinity-dependent exchange of antigen)
  • interaction with follicular T helper cells (including survival signaling and rescue)
  • affinity maturation as an emergent phenomenon

More importantly, however, PlazaSur highlighted the limitations of the traditional approach to simulating biological processes and motivated an innovative approach that was required of the second-generation software platform (Daphne).

In the traditional approach, as exemplified by PlazaSur, complex biological processes are conceived as biological models, the biological models are formulated as mathematical models, and these mathematical models are encoded as software (Figure 1a). There are two inherent problems with this approach: once compiled, the user is constrained by the encoded biological model and only has control over parameter values and the hard-coded biological model cannot be easily updated to accommodate new phenomena.

Therefore, the major motivations for the design of Daphne were the needs for the user to have greater flexibility in defining the biological model and to be able to continually refine the model as new data and new information are acquired. Building on our experience putting together the special-purpose germinal center simulation platform PlazaSur, we settled on a set of design principles that will permit the greatest degree of flexibility to the user, while maintaining the power to run powerful, realistic simulations.

There are two opposing needs that must be resolved. First, the user must have enough flexibility to specify a model sufficient to the task of representing the system of interest. Second, the universe of models under consideration must have sufficient structure that the platform can construct any model in the class at run-time, and can do so enabling high-performance execution of even quite complex models.

To accomplish this, we move from an object-centric to a process-based approach. All the dynamics, except diffusion and cell collisions, are implemented as reactions. This permits modular modeling, in which coarse-grained models can be replaced by more detailed models as the need arises. This approach also provides greater flexibility for the user – new molecules, reactions, force-molecular field relationships, and cell phenotypes can be added in a plug-and-play manner at runtime without the need to recompile (Figure 1b). Finally, this approach provides the capability for the code to remain relevant in the rapidly-changing field of immunology by providing the capability to incorporate new phenomena as they are revealed by future experiments. On-the-fly adjustment of resolution level is particularly important as we begin the modeling of B-cell activation and the complex molecular dynamics within the FDC-B-cell synapse.

_images/daphne_overview.png

Figure 1: The traditional approach to modeling biological processes, as exemplified by the first-generation software, PlazaSur. The reaction-based approach of the second generation software, Daphne.

In summary, Daphne is built on the generic processes of Newtonian mechanics, and generalized chemical kinetics, as well as a small number of cell biology-specific constructs. The Daphne system simulates in continuous time and continuous space, and uses mass-action kinetic ordinary differential equations realized in a moment expansion for cytoplasmic biochemistry and plasma membrane biochemistry, reaction-diffusion partial differential equations for extracellular soluble factors, Markov models for discrete cellular transition such as cell division, differentiation and death, and stochastic processes for cellular chemokine sensing and locomotion.

The user has complete control over the chemistry and cell biology in any given simulation. He or she may run simulations as provided, alter reaction rates or force parameters, or start from scratch and build a complete chemistry, along with novel cells constructed using the new chemistry.