Key Decisions for the Development of the Next-Generation of Context-Specific Models

Genome-scale metabolic models provide a valuable context for analyzing data from diverse high-throughput experimental techniques. Since some enzymes are only active in specific environments, several algorithms have been developed to build context-specific models. However, the content and associated predictive capacity of resulting models is considerably impacted by the different assumptions used at each step of the extraction process: from the decisions on how to overlay data onto networks to the parameter choices for extraction algorithms^1,2. These choices lead to a poor consensus in generated models, which may limit the use of context-specific methods for data-driven hypothesis.

We present a comparative analysis of existing extraction algorithms and an assessment of the key decisions influencing the data contextualization methods. From this work, we propose an approach to obtain better consensus across existing extraction algorithms and more accurate models. This approach is enabled with a framework we built for inferring metabolic functions that should be active in a specific context directly from transcriptomic data. These functions can be used to protect associated biochemical reactions during the implementation of extraction methods. The promising results obtained using this protectionist approach underline the potential interest of describing genome-scale metabolic reconstructions as more than a network of reactions but rather as an interconnected map of cellular functionalities.

1. Opdam, S., Richelle, A., Kellman, B., Li, S., Zielinski, D. C., & Lewis, N. E. (2017). A Systematic Evaluation of Methods for Tailoring Genome-Scale Metabolic Models. Cell Systems, 4(3). http://doi.org/10.1016/j.cels.2017.01.010

2. Richelle, A., Joshi, C., & Lewis, N. E. (2018). Assessing key decisions for transcriptomic data integration in biochemical networks. BioRxiv, 301945. http://doi.org/10.1101/301945