A combinatorial view of canonical Wnt signaling

Abstract

This computational thesis is concerned with two problems, one methodological, the other scientific. The scientific problem consists in understanding the behavior of the canonical Wnt signaling pathway at a molecular systems level. Several components of the pathway are scaffold proteins, capable of polymerization. This is the first model of Wnt signaling that has sought to elucidate the role of polymerization and scaffold use, notably regarding complex formation and combinatorial post-translational modifications. As for the methodological problem, I wanted to build a model based directly on a collection of mechanistic facts derived from the primary research literature, without this collection organized by some a priori understanding, or prejudice, of how the pieces fit together. The model was built in a haphazard and incremental fashion, along the way certifying its compliance with experimental observations. Its unbiased nature promoted unanticipated insights. This thesis is therefore a contribution to a new style of modeling. A technical problem consisted in identifying a platform that would permit this task. Although one such existed, it was in a testing stage. By breaking, debugging, and exploiting it many times, I contributed to its development. A further methodological problem consisted in the analysis of a model of this nature. A large model, in which behavior emerges from the asynchronous interaction of components is effectively "wild". Building it does not yield automatic understanding of it. Insight required an approach that cannot be described in any other term than "empirical". Tools suitable for its analysis did not exist, and I had to build them. This also led to the making of a model’s model, in which "empirical" insights from the wild and "undomesticated" model were marshaled into a more traditional analytical one. The insights that emerged from the wild model are: (i) polymerizing scaffolds can structure larger programmable entities, enabling concentration effects through uni-molecular interactions; (ii) the β-catenin destruction complex is not a classical complex with fixed stoichiometry, but a statistical machine; (iii) the Wnt signalosome is not an entity distinct from the destruction complex it antagonizes; rather one complex literally morphs into the other, never completely abolishing destruction activity.