We develop criteria based on a calibration argument via discrete PDE and semidiscrete optimal transport, for finding sharp isoperimetric inequalities of the form $(\sharp Ω)^{d-1} \le C (\sharp \overrightarrow{\partialΩ})^d$ where $Ω$ is a subset of vertices of a graph and $\overrightarrow{\partialΩ}$ is the oriented edge-boundary of $Ω$, as well as the optimum isoperimetric shapes $Ω$. The method is a discrete counterpart to Optimal Transport and ABP method proofs valid in the continuum, and answers a question appearing in Hamamuki \cite{hamamuki}, extending that work valid for rectangular grids, to a larger class of graphs, including graphs dual to simplicial meshes of equal volume. We also connect the problem to the theory Voronoi tessellations and of Aleksandrov solutions from semidiscrete optimal transport. The role of the geometric-arithmetic inequality that was used in previous works in the continuum case and in the $\mathbb Z^d$-graph case is now played by a geometric cell-optimization constant, where the optimization problem is like in Minkowski's proof of his classical theorem for convex polyhedra. Finally, we study the optimal constant in the related discrete Neumann boundary problem, and present a series of possible directions for a further classification of discrete edge-isoperimetric constants and shapes.