Precision in Gene Regulation
Current work on this topic focuses on investigating both precise sensing schemes and information-theoretic inference schemes. In this context, we are interested in optimal network architectures or optimal decision pathways, also more abstractly. On the biological side, we are interested in different enhancer architectures, as well as comparing these architectures with models.
The expression of genes needs to be regulated precisely, especially when these genes encode proteins that are important for an organisms to develop in a healthy way. However, genes are regulated by transcription factors or other proteins which are frequently expressed in reasonably low numbers; for the fly morphogen Bicoid, these numbers range between 10000 molecules per cell in the anterior to of order 10 molecules at the posterior side of the fly embryo. In addition, these molecules may interact with each other while traveling to the region on the genome where they can regulate a specific gene; recently, this clustering of molecules has been investigated given the increasing interest in liquid-liquid phase separation. These complexities suggest that molecular signals can be noisy, and that organisms need to have found a way to nevertheless respond to the signals precisely. Advances in high-resolution, single molecule experiments, also regarding the spatial patterns of gene expression or spatial clustering of certain proteins, or in the specific manipulation of genomic regions in both in-vivo and synthetic system make it possible to address this conundrum now: theory towards understanding regulatory precision can now be based on experimental data. Elucidating the regulatory logic and kinetic or thermodynamic tricks that contribute to regulatory precision is one of the main goals of this group.
Relevant Publications:
M. Bauer, M. Petkova, T. Gregor, E.F. Wieschaus and W. Bialek,
Proceedings of the National Academy of Sciences 118 (46), e2109011118 (2021).
In this work, we analyze network of early transcription factors in the fly embryo, which are responsible for the healthy body part segmentation along the head-tail axis. We used an information-theoretic formulation to infer an optimal sensor for these transcription factor concentration profiles, in order to see if the fly's enhancers that perform this sensing have notable features in common with an optimal sensor. Indeed, on an abstract level, this seems to be the case: we found that one would need many sensors, not just one, and that sensors should not measure just one of these transcription factors each. Thus, the fact that many enhancers have a combination of binding sites for several of these transcription factors is consistent with information-theoretic optimality for sensing. This finding was also encouraging because it shows that we can learn from model-free calculations.
