Composition and Quantitative Prediction of Transcriptional Circuits in Mammalian CellsView all posters
A long-standing goal of synthetic biology is to rapidly engineer new regulatory circuits from simpler regulatory elements whose properties have previously been characterized individually. A critical impediment, however, has been the lack of accuracy in predictions of circuit behavior made by computational models. The typical constructive approach of synthetic biology has led to models that explicitly encode all the biochemical reactions believed to be significant. Such models, however, generally rely on difficult to estimate parameters; moreover, while the impact of cellular context on regulatory motifs and signaling pathways is becoming increasingly clear, this impact is not sufficiently well understood to incorporate effectively in reaction models. Here we introduce a new method, Empirical Quantitative Incremental Prediction (EQuIP), which addresses these problems and provides accurate predictions of biological circuit behavior. In EQuIP, the basic unit of a circuit is a “device” encapsulating a set of regulatory interactions. For each device, a composable model of its incremental input/output expression phenotype is derived solely from empirical observation of expression dynamics and device steady-state behavior in the cellular context. This approach both abstracts away biochemical details and captures significant interactions with cellular context without requiring either explicit or well-understood models of those interactions. We validate EQuIP by characterizing three transcriptional repression devices with transient transfection in mammalian HEK293FT cells, then precisely predict the behavior of six cascades, each comprising two of these repressors. For a range of induction levels, these cascades exhibit up to 18-fold +/- difference in fluorescence and over 1000-fold cell-cell variation in fluorescence, yet EQuIP’s computational predictions have a mean error of only 1.6-fold compared to experimental data. Such accurate predictions will allow synthetic biologists to determine combinations of devices likely to produce a desired behavior, thereby allowing reliable forward engineering of complex biological circuits from libraries of standardized devices.