This was a super fun and interdisciplinary project that couldn't have been done without my amazing co-authors, my fantastic advisor Michael, and everyone else from the @elowitzlab.bsky.social!

We think our results and quantitative framing of protein interaction networks will help us better understand natural signaling pathways - as well as engineer powerful new types of synthetic circuits! (stay tuned...)

Lastly, we demonstrate that even networks with RANDOM dimerization affinities, when large enough, can be tuned to perform nearly ANY input-output function simply by adjusting the protein expression levels

We demonstrate that dimerization networks can readily compute complex three- and four-input logic gates, like the complex "exactly any 2 or None" gate shown here

Further, the SAME set of proteins can compute DIFFERENT functions in different cell types, simply by expressing the protein components at different levels

Using a computational model, we first found that dimerization networks can compute a wide variety of complex and multi-input functions

We lacked a detailed understanding of dimerization network computation: What types of computations are/aren't possible? Can these networks do computations on multiple inputs? What about cell-type-specific computations? How big do these networks need to be?

These dimerization networks act as biochemical computers: Input signals regulate individual monomers, which then form output dimers that regulate downstream gene expression.

The motif of "many-to-many" binding interactions can be found at all levels of cell signaling. Here we focused on systems in which multiple monomers competitively dimerize in different combinations, such as in the bHLH and bZIP transcription factors.

How can different cell types use protein interaction networks to make decisions based on multiple signals?