Machine learning opens a doorway for microrheology with optical tweezers in living systems
Matthew G. Smith,Jack Radford,Eky Febrianto,Jorge Ramírez,Helen O’Mahony,Andrew B. Matheson,Graham M. Gibson,Daniele Faccio,Manlio Tassieri
DOI: https://doi.org/10.1063/5.0161014
IF: 1.697
2023-07-01
AIP Advances
Abstract:It has been argued that linear microrheology with optical tweezers (MOT) of living systems “is not an option” because of the wide gap between the observation time required to collect statistically valid data and the mutational times of the organisms under study. Here, we have explored modern machine learning (ML) methods to reduce the duration of MOT measurements from tens of minutes down to one second by focusing on the analysis of computer simulated experiments. For the first time in the literature, we explicate the relationship between the required duration of MOT measurements (Tm) and the fluid relative viscosity (ηr) to achieve an uncertainty as low as 1% by means of conventional analytical methods, i.e., Tm≅17ηr3 minutes, thus revealing why conventional MOT measurements commonly underestimate the materials’ viscoelastic properties, especially in the case of high viscous fluids or soft-solids. Finally, by means of real experimental data, we have developed and corroborated an ML algorithm to determine the viscosity of Newtonian fluids from trajectories of only one second in duration, yet capable of returning viscosity values carrying an error as low as ∼0.3% at best, hence opening a doorway for MOT in living systems.
materials science, multidisciplinary,physics, applied,nanoscience & nanotechnology