Speaker
Description
Cell movement is an important part of many biological processes, such as collective cell migration and the immune response to cancer. In many systems interactions between nearby cells are a key driver of cell movement, yet it can be difficult to infer and express the rules governing cell-cell interactions in a manner that is biologically interpretable. Here we present a model, based on the theory of deep attention networks, that learns how cell-cell interactions affect cell movement from cell trajectory data. We develop a suite of tools that leverage the structure of this model to present the learned interaction dynamics in an interpretable manner. Our work extends previous applications of deep attention networks to cell movement, moving beyond only inferring how strongly cells interact in monocultures, to also inferring the effect of these interactions on cell movement in both monotype and multi-type cell movement systems.
We use our model to understand how cytotoxic T cells and tumour cells interact in an in vitro co-culture, and use these results to build an agent-based model of the T cell response to cancer. Recent experimental studies have observed significant variability in the killing capabilities of cytotoxic T cells – i.e. some T cells kill many more cancer cells than others. We use our model to investigate what mechanisms might give rise to this heterogeneity, and ask whether it can be explained by a model in which all T cells follow the same behavioural rules.
