(622g) Application of Machine Learning to Accelerate High-Throughput Molecular Dynamics Screening: A Study of Tribological Properties of Monolayer Films

Monolayer films have shown promise to reduce friction and wear of mechanical devices with separations on the nanoscale.¹ These films have a vast design space owing to the many tunable properties that have been shown to impact tribological behavior, including terminal group chemistry, film thickness and film composition. Here, the Molecular Simulation Design Framework (MoSDeF)² is used to perform a combinatorial screening study of 9747 unique multi-component monolayer films (116,964 total simulations) to identify which terminal groups combinations and compositions lead to favorable tribological properties, i.e., those that minimize the coefficient of friction and adhesion forces. A random forest regressor machine learning framework³ is developed and used to establish a more robust understanding of the connections between terminal group chemistry and tribological properties. The impact on the predictive capacity of the size/scope of the training set used to develop the machine learning models is also investigated. Here we present heuristics for determining the accuracy and efficiency of ML models for use in prescreening. The machine learning model was subsequently applied to screen 193,131 unique film candidates identified from the ChEMBL small molecule library, resulting in approximately a 5 order of magnitude speed-up in analysis compared to simulation alone. This work demonstrates the promise of computational screening, in combination with machine learning, to greatly increase the throughput in combinatorial approaches.⁴ The simulations reported follow guidelines suggested by the TRUE standard, emphasizing the reproducibility and extensibility of the study.⁵

References

[1] N. S. Tambe and B. Bhushan, “Nanotribological characterization of self-assembled monolayers deposited on silicon and aluminium substrates”, Nanotechnology, 2005, 16, 1549–1558.

[2] “MoSDeF” [Online]. Available: https://mosdef.org

[3] A. Z. Summers, J. B. Gilmer, C. R. Iacovella, P. T. Cummings and C. McCabe, “MoSDeF, a Python Framework Enabling Large-Scale Computational Screening of Soft Matter: Application to Chemistry-Property Relationships in Lubricating Monolayer Films”, J. Chem. Theory Comput., 2020, 16, 1779–1793.

[4] C. D. Quach, J. B. Gilmer, D. Pert, A. Mason-Hogans, C. R. Iacovella, P. T. Cummings and C. McCabe, "High-Throughput Screening of Tribological Properties of Monolayer Films using Molecular Dynamics and Machine Learning", J. Chem. Phys. 2022, in press 10.1063/5.0080838.

[5] M. W. Thompson, J. B. Gilmer, R. A. Matsumoto, C. D. Quach, P. Shamaprasad, A. H. Yang, C. R. Iacovella, C. McCabe and P. T. Cummings, “Towards molecular simulations that are transparent, reproducible, usable by others, and extensible (TRUE)”, Mol. Phys., 2020, 0, e1742938.