Title: Software infrastructure for simulations of cellular-scale biology
Abstract: Cellular biology is rife with problems that are difficult to study in a purely experimental context. Perturbing a protein sequence, changing the cellular environment, or altering mechanical properties often disrupts many coupled, redundant cellular processes at once, which in many cases is fatal to the cell under study. Modeling and simulation can offer a solution to this problem; if a simulation captures the relevant biology, it can be used as a facsimile to study the underlying system. Many seemingly disparate systems share the shape of their components: objects that diffuse and move, grow and shrink, and bind and unbind. Structural similarities beget computational similarities, but the complexity of these objects and interactions restricts the set of applicable tools. For instance, at the atomistic scale of molecular dynamics, many tools exist for treating objects as point-like particles. However, for systems at the level of the cellular cytoskeleton, few general-purpose libraries exist; most are research codes that are not performant, well tested, or extensible. Here at the Flatiron Institute, we have the capabilities and long-term stability to create and maintain infrastructure capable of addressing such critical gaps. In this talk, I present MuNDy (multibody nonlocal dynamics), a flexible multi-GPU dynamics engine for building large-scale simulations of diverse cellular-scale problems ranging from bacterial colonies to the mitotic spindle.