Title: Integrating diverse analyses reveals a convergent design principle of B cell differentiation
Speaker: Tongli Zhang (Applied Math)
Abstract: All cells are information processing systems. In response to environmental signals such as pathogen invasion, cells such as. B cells execute proper responses that are governed by their gene regulatory network (GRN). Upon activation, B cells takes an exceptional developmental dynamic that first bifurcates and then converges to a singular state of plasma cells. A complex GRN, comprising of sequential double negative feedback loops that are connected by both coherent and incoherent feedforward loops, has been proposed to account for this unusual trajectory of B cells. As it is straightforward to translate this GRN into differential equations, the unknown values of parameters proposes a fundamental challenge to understand this system and design optimal strategies for further study. To address this challenge, we have translated the known network into mathematical equations and reverse engineered the values of parameters. After multiple models representing differentiating B cells are selected, we extracted unbiased trends from the populations by systematic application of machine learning methods. By revealing the most influential control parameters, machine learning suggested the candidates that might play the major roles in B cell differentiation. Furthermore, it also suggested that cells of different fates can be well separated on the basis of few key parameters. To achieve mechanistic insights, we also carried out nonlinear dynamical analysis. Our analysis demonstrates that the exceptional properties of differentiating B cells are manifested in a restricted parameter space, which lies near the bistability region. Consistent from the results of machine learning, this conclusion provides a novel insight into the design principle of B cell differentiation. Furthermore, our work demonstrates that a novel and creative combination of diverse methodologies can be used to extract convergent and deep understanding of the complex biological control network even though our knowledge of the network is incomplete.