The most routinely synthesized molecular motors of the day are not nearly as sophisticated as their biological counterparts in living cells. Despite several milestone achievements, including the Nobel Prize of 2016, there remains major bottlenecks in the conceptualization of artificial molecular motors, tuning their energy efficiencies, and optimizing their turnover rates. The biggest challenge is our incapability of finding a library of molecular components that assemble into a desired motor. Here, combining rate-kinetic models, molecular simulations on petascale to exascale supercomputers, and machine-learning approaches, we will develop a high-throughput search algorithm, which based on energy changes, predicts sets of motor action-friendly chemical components, and their designs into de novo molecular motors.Caption. Parallels between the biological (left) and abiological (right) motors that MotorBuilder will exploit to predict bioinspired designs of artificial molecular motors.
We propose that the high-throughput characteristics of biological designs exemplified by evolution can be mimicked using computational approaches to accelerate the engineering of functionally-efficient artificial motors. Combining molecular simulations and machine-learning approaches on exascale supercomputers, we will develop an integrated search and learning tool, MotorBuilder, to explore the space of perspective molecules for porting into a desired motor-action design; sampling of such high-dimensional spaces is manually implausible. Mirroring the evolutionary search of designs in biology, the diversity of chemical componentry identified by our computational search will manifest either in improved designs of existing motors or overcome assembly limitations to produce molecular motors de novo. MotorBuilder will be tested with two different linear motors and one rotatory motor. The computational predictions will be experimentally verified with long-term collaborator A. Flood (http://www.indiana.edu/~floodweb/).