Error-rate reduction in network-based biocomputation

TL;DR

This paper optimizes the geometry of pass junctions in network-based biocomputation to significantly reduce error rates, enabling larger and more scalable molecular computing networks with potential for further improvements using 3D structures.

Contribution

It introduces a systematic approach to design and experimentally validate low-error pass junctions in NBC, achieving under 1% error rate and highlighting the need for 3D junctions for further reduction.

Findings

Optimized pass junction design reduced error rate to below 1%.

Experimental and simulation results aligned better with a layer of myosin included.

Further error reduction likely requires three-dimensional junctions.

Abstract

Network-based biocomputation (NBC) is an alternative parallel computing paradigm that encodes combinatorial problems into a nanofabricated device's graphical network of channels, enabling cytoskeletal filaments propelled by molecular motors to explore the problems' solution space. NBC promises to require significantly less energy than traditional computers due to the high energy efficiency of molecular motors. However, error rates associated with the pass junction crossing, the primary path-regulating geometry, pose a bottleneck for scaling up this technology. Here, we optimize the geometry of the pass junction for low error rates for the actin-myosin system. To do so, we evaluate various pass junction designs that differ in features, such as the nanochannel width, junction crossing area, and angles of a funnel-shaped output part of the junction. Error rates were measured experimentally by using gliding motility assay and as well as by simulation methods. The final optimized design displayed a decreased error rate of under 1 percent compared to the previous 2-4 percent. We anticipate this improvement will enable scaling up NBC networks from tens to hundreds of pass junctions. However, the results of 2D junction optimizations also suggest that further drastic reduction of error rates in two-dimensional pass junctions is unlikely, necessitating three-dimensional junctions, such as bridges or tunnels, for complete error rate mitigation. Furthermore, the simulation results demonstrate that including a layer of myosin motor on the channel provides a better fit between simulation and experimental results.