Molecular cascades introduced in provide new ways to exploit the motion of individual molecules in nanometer-scale structures. Computation is performed by purely mechanical means similarly to the toppling of a row of standing domino. A specific feature of molecular cascades is that an inverter cannot be build, because it would require that all molecules in the inverter's output untopple when the input cascade topples. This is not possible because
an untoppled state has higher energy than a toppled one. As a solution, we propose to avoid the need for inverters by representing
signals by the dual-rail convention. As a basic building block we use
a molecular block, which has four inputs x1,...,x4 such that x3 = x'1, x4 = x' x2, and two outputs ƒ1 = x1 • x2
and ƒ2 = x3 + x4. If input variables are available in both complemented and non-complemented form, then any Boolean function can be implemented by a composition of such molecular blocks. We present an experimental tool which first uses a rule-based randomized search to optimize a Boolean network and then maps it into a network of interconnected molecular blocks.