The modeling of transcription to genome proximal elements to date has revealed associations, but seldom are disruptions performed to confirm mechanistic possibilities and substantiate causality. In studying multistate cell systems, many processes important to human health are difficult to study due to low cell availability and/or dyssynchrony leading to heterogeneous cell populations. We propose to study a model of the human epidermal differentiation system, which by its intrinsic properties does not have these problems but at the same time closely simulates the native process. We plan to perform multiple next generation sequencing modalities of transcription and gene proximal components over a time course spanning the transition from progenitor to differentiated keratinocytes. A network based on boosting methods and dynamic Bayesian networks will then be generated to model transcription to the gene proximal components, and this construct will be tested with various regulatory disruptions. Moreover, additional assays of transcription and gene proximal components will be performed during intervals which the model suggests will be particularly illuminating for epidermal differentiation. With these new data, the model will be further refined, and this cycle will be repeated multiple times. Because of the tractability of our experimental system to a vast array of sequencing assays and regulatory disruptions, we will be able to achieve an understanding of how much each assay contributes to the predictive accuracy of our model. In this way, our results will have implications not only for skin biology and hundreds of skin disorders but also for network modeling of transcriptional regulation in general.
The regulome of cells is, to date, incompletely understood. We plan to perform multiple next generation sequencing modalities on an epidermal differentiation system over a time course. A network model will then be constructed and tested with various regulatory disruptions. Because of the tractability of our experimental system to sequencing assays and regulatory disruptions due to the ease of biomaterial generation, we will be able to achieve an understanding of how much each assay contributes to the predictive accuracy of our model. In this way, our results will have implications not only for dermatology but also for network modeling of regulation in general, especially for systems which are more difficult to study experimentally due to low cell availability, etc.
