The modeling of biological processes is difficult in part because of the complex actions at different levels within the organism. The authors present a modeling approach that operates on two scales of action. Of special interest is the use of agent-based modeling to reflect the stochastic nature of portions of the biological process. They demonstrate the concept on the problem of understanding osteoporosis, especially its progress related to the proteins RANK and RANKL.
Bones are constantly renewed by the absorption of older--and the regeneration of newer replacement--tissues. Imbalance (for a variety of reasons) can cause bone loss; a common mode of this imbalance is the condition known as osteoporosis. Pairs of osteoclasts (which destroy bone) and osteoblasts (which produce bone) carry out this process; the authors present a shape calculus to describe the actions of these cells. Some enhancements to their previously published description of the shape calculus (see reference [3] in the paper) were done to better model features of the biological process. Events in time such as splits and recombinations of the osteoblast and osteoclast pairs are included. The resulting description has parameters set to best describe normal bone growth at a collective level.
However, accurate modeling of this process also required inclusion of some stochastic elements. Rather than directly adding stochastic elements to their shape calculus model, the authors opted to define an agent-based model operative at a lower level. This is the multiscale viewpoint mentioned in the title. The agent-based level is strongly influenced by the spatial relationships of the agents to mimic the importance of proximity in the action of the cells.
The authors show results of their model simulating bone regrowth of a microfracture. One clear outcome is the dependence on the signaling activity of the RANKL protein, confirming the model’s representation of the effects of osteoporosis. Long-term differences in bone density related to osteoporosis are well predicted by the two-scale model. The authors mention future extensions of this interesting work. Most exciting is the statement that they propose to extend the model to comprehend the intracellular pathway, allowing modeling of the effects of drug therapy.
This paper is generally well written and quite approachable for those with less familiarity with the techniques used. The fundamental approach of using a multiscale model to include stochastic effects at one level is interesting and likely has applications to other processes that are complex and difficult to model.