New technique uses artificial intelligence to calibrate force fields
Physicists at the University of Gothenburg have developed a computer program that effectively measures the forces exerted on microscopic particles. This is of importance for our understanding of the structure and function of particles..
Measuring the forces that are exerted upon microscopic particles is of the utmost importance for understanding both the particles’ structure and function. The method can be applied to calculate a cell’s elasticity, to understand how a virus moves through the air or how microscopic particles function.
The force fields exerted upon microscopic particles are calibrated through an analysis of their location history. This was previously done using mathematical formulas. However, it is now possible to use artificial intelligence to analyse a particle’s location history.
“By using advanced machine learning methods, we have demonstrated not only that these measurements can be done with greater precision, but that they also can be applied in a greater number of cases,” says Aykut Argun, doctoral student at the Department of Physics at the University of Gothenburg.
Open source software programme
The researchers have also made their computer software program DeepCalib available as open source code so that their method can be used widely.
The program, which can be optimised for specific force fields, is based on recurrent neural networks that address the data pattern which changes over time based on continuous feedback. In this way, they can learn from past occurrences and use them to predict the future.
The researchers are now planning to expand DeepCalib’s functions to cover more complex biological systems, such as proteins moving through blood.
The results were recently published in Applied Physics Reviews. They were also showcased on the American Institute of Physics’ Scilight webpage, which presents the most interesting current research in the field of physics.
The article in Applied Physics Reviews can be accessed here: https://aip.scitation.org/doi/10.1063/10.0002653
Scilight: https://aip.scitation.org/doi/10.1063/10.0002653
Contact:
Aykut Argun, doctoral student, Department of Physics, University of Gothenburg, aykut.argun@physics.gu.se +46-(0)31-786 91 57, +46-(0)766-22 91 57
Photo: Aykut Argun, photographer Alessandro Magazzu.
Title: Enhanced force-field calibration via machine learning
Link: https://aip.scitation.org/doi/10.1063/5.0019105
Article in Applied Physics Reviews: https://aip.scitation.org/doi/10.1063/10.0002653