MEASUREMENT ERROR CORRECTION IN PARTICLE TRACKING MICRORHEOLOGY

• 成果类型: Article
• 署名作者: Ling, Yun; Lysy, Martin; Seim, Ian; Newby, Jay; Hill, David B.; Cribb, Jeremy; Forest, M. Gregory
• 署名单位: University of Waterloo; University of North Carolina; University of North Carolina Chapel Hill; University of Alberta; University of North Carolina; University of North Carolina Chapel Hill; University of North Carolina; University of North Carolina Chapel Hill; University of North Carolina; University of North Carolina Chapel Hill
• 刊物名称: ANNALS OF APPLIED STATISTICS
• ISSN/ISSBN: 1932-6157
• DOI: 10.1214/21-AOAS1565
• 发表日期: 2022
• 页码: 1747-1773
• 关键词: single-molecule tracking; living cells; subdiffusion; transport; motion; nanoparticles; diffusion; MODEL; FLOW

摘要

In diverse biological applications, single-particle tracking (SPT) of passive microscopic species has become the experimental measurement of choice, when either the materials are of limited volume or so soft as to deform uncontrollably when manipulated by traditional instruments. In a wide range of SPT experiments, a ubiquitous finding is that of long-range dependence in the particles' motion. This is characterized by a power-law signature in the mean squared displacement (MSD) of particle positions as a function of time, the parameters of which reveal valuable information about the viscous and elastic properties of various biomaterials. However, MSD measurements are typically contaminated by complex and interacting sources of instrumental noise. As these often affect the high-frequency bandwidth to which MSD estimates are particularly sensitive, inadequate error correction can lead to severe bias in power law estimation and, thereby, the inferred viscoelastic properties. In this article we propose a novel strategy to filter high-frequency noise from SPT measurements. Our filters are shown theoretically to cover a broad spectrum of high-frequency noises and lead to a parametric estimator of MSD power-law coefficients for which an efficient computational implementation is presented. Based on numerous analyses of experimental and simulated data, results suggest our methods perform very well compared to other denoising procedures.