Extended Data Fig. 3: Suppression of chimeric growth through tile set design.

a, We use simulations to contrast assembly errors in three distinct tile sets: the proofreading tile set with an inert boundary used in experiments, described in Fig. 2(a, top); a simple checkerboard tile set with a strictly alternating shared and unique tile pattern for each shape, where unique tiles can be seen as mediating different interactions between shared tiles (a, middle); and an edge-guarded checkerboard in which we additionally enforce inert bonds around each shape’s perimeter (a, bottom). For each tile set, we performed kinetic growth simulations, starting from a pre-formed 5 × 5 seed taken from a location within H. Simulations were performed using the kinetic Tile Assembly Model as implemented by xgrow (with chunk fission)⁶² with uniform tile concentrations corresponding to 62 nM and parameters estimated in Supplementary Information section 2.1. b, Schematic illustrates various desired and undesired growth pathways for A, along with representative AFM images taken from the A flag 1 experiment (Supplementary Information section 5.3.13). Two distinct kinds of chimeric structures were seen in simulation as the result of promiscuous interactions: chimeric structures can grow either before full assembly of the target structure (e.g., part-A, part-M) or emerge spontaneously from the edge of a properly formed structure (e.g. full-A, part-H). Chimeras like those illustrated along the lower path are held together by just a few bonds and sometimes can quickly break apart (tiles with unintended bonds are shown in red); these result in sharp drops in simulated assembly size, as the simulation discards one subassembly when disconnected. Note that chimeric growth was not observed experimentally, possibly as a result of effective experimental system design; however, many observed structures failed to complete the upper right and/or lower left corners, or appeared to have suffered a spiral growth defect. A possible explanation for the missing corners, which is also seen in H and M, is supported by coarse-grained molecular dynamics simulations of SST lattice curvature (Supplementary Information section 3.4). Spiral defects were not seen in H or M and are presumably due to the interior hole in A. c–e, The size of the assembly (in units of the size of the fully formed H) is shown as a function of time. For higher temperature 48.9 °C (c), no chimeras are observed on the simulated timescales for any tile set. For intermediate temperature 47.2 °C (d), all 6 checkerboard trajectories still result in chimeras, while no errors are observed on the timescale probed for the guarded checkerboard or experimentally-implemented proofreading tile set. For lower temperature 45.5 °C (e), chimeras are seen in all runs for checkerboard structures (red traces), 4 of the 6 runs for guarded checkerboard structures (green traces) and 1 of the 6 runs for proofreading structures.