Efficiently and effectively recognizing toricity of steady state varieties (Q2035620)

This paper applies methods of computational algebra to the study of models describing biochemical reaction networks. The main theoretical problem addressed is how to decide, given a polynomial ideal \(I\) in \(n\) variables, if it defines a \textit{shifted toric variety} -- that is, whether or not the set of solutions with all coordinates nonzero \(V(I)^*\) is a coset of some subgroup of the \(n\)-dimensional standard algebraic torus over \(\mathbb{C}\) or \(\mathbb{R}.\) When \(V(I)^*\) is a subgroup, the variety is said to be toric, in agreement with the usual notion. Over \(\mathbb{C}\), there is a fairly simple criterion: Proposition 7 states that any reduced Gröbner basis of the ideal \(\sqrt{I : \langle x_1 \cdots x_n \rangle }\) must solely consist of binomials of the form \(c_{\alpha } x^\alpha - c_{\beta} x^\beta \) where the coefficients \(c_\alpha , c_\beta \) are nonzero. In the special case where all coeffcients are \(1,\) the nonzero solutions form a subgroup. These observations lead to an effective test for (shifted) toricity. The authors use Maple to compute primary decompositions of several \emph{steady-state} models from systems biology and apply this test to each resulting component \(Q\) after eliminating variables that vanish at all points in \(Q.\) Experimentally, the authors find that a majority of cases are either \(V(Q)^* = \emptyset \) or the variety is shifted toric. The authors then proceed to give a quantifier-elimination based criterion for (shifted) toricity over \(\mathbb{R}\) in the form of \textbf{Algorithm 5} \(\operatorname{Classify}_{\mathbb{R}}\). They make similar experimental observations applying this criterion with the QE-based system \texttt{REDUCE} to slightly smaller subset of the data used over \(\mathbb{C}.\) Finally, a theoretical complexity analyses for various procedures and a detailed appendix describing experimental setup are presented.