Tl;dr per my read 1) Convert the graph representation to a biological encoding that leverages DNA's complement binding. 2) Shake 3) Read unique newly bound s

Tl;dr per my read 1) Convert the graph representation to a biological encoding that leverages DNA's complement binding. 2) Shake 3) Read unique newly bound sequences that contain your desired properties (e.g. start at X, end at Y), which will by chemical properties include all possible answers with high probability The encoding seemed the tricky / novel part.