Using Qualitative Data for Quantitative Dynamics Estimation in Systems Biology

Binary biological transitions are prevalent in microbial systems, ranging from the reversal of mobile genetic elements’ persistence to the change of complex species’ interaction type. While these binary data are often generated, they are rarely used to assist quantitative dynamics information prediction. We investigate whether such binary qualitative data can be used to estimate quantitative data in these systems. We first demonstrate an approach to predict both qualitative and quantitative systems outcomes with mainly qualitative data via a “shape + structure” model. In this approach, we use machine learning to predict the boundary of the transition with qualitative data. This boundary is then used to create contours, the shape, of estimated quantitative change. Some quantitative data points are used to provide additional information, the structure, of the real quantitative outcome. We use this structure to select the best machine learning model for both qualitative and quantitative predictions. To illustrate this approach, we predict the plasmid qualitative persistence and quantitative abundance information using mainly persistence data in a simple simulated community. We also show that plasmid dynamics in more complicated simulated communities and real experimental communities can be estimated in the same manner. Beyond this system, we demonstrate that a wide range of biology systems have similar property where qualitative information can be used to infer quantitative information. Our results indict the underused value of qualitative data in quantitative dynamics prediction in microbial communities.