What can, and cannot, be predicted

One question that pervades the study of complex systems is our capacity to predict its function. We are starting to think about this issue in recent years. One common solution to this issue is to measure, and measure everything. When talking about cell physiology this implies to quantify gene expression, metabolites, etc. When talking about behavior, this implies to measure neuronal activity. Say, we can develop this approach. Will it increase predictability? In the cellular context, recent coarse models are starting to provide an alternative to this program by providing valid predictions. We recently examined that in the context of regulation. We also encounter this question when trying to predict the function of artificial microbial consortia. We found that function relies on higher-order interactions and that this restricts our capacity to predict. This problem also takes us to the issue of the influence of genetic background in the prediction of the consequence of mutations, with practical implications, i.e., personalized medicine. If theory is not possible, should we trust instead predictions that rely on automatic systems?

Some readings

Genetic buffering and potentiation in metabolism (2019). [HTML].

Complex genetic and epigenetic regulation deviates gene expression from a unifying global transcriptional program. With M. Chagoyen (2018). [HTML].

High-order interactions dominate the functional landscape of microbial consortia. With D. Bajic, M.L. Osborne, A. Sanchez and A. Sanchez-Gorostiaga (2018). [Preprint] and [HTML].