Considering experimental frame rates and robust segmentation analysis of piecewise-linear microparticle trajectories
Keisha J. Cook \corrauth, Nathan Rayens, Linh Do, Christine K. Payne, and Scott A. McKinley · 2024
Read Considering experimental frame rates and robust segmentation analysis of piecewise-linear microparticle trajectories by Keisha J. Cook \corrauth, Nathan Rayens, Linh Do, Christine K. Payne, and Scott A. McKinley (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-21025.txt.
Description
Abstract The movement of intracellular cargo transported by molecular motors is commonly marked by switches between directed motion and stationary pauses. The predominant measure for assessing movement is effective diffusivity, which predicts the mean-squared displacement of particles over long time scales.
Catalog record
| CREATOR | Keisha J. Cook \corrauth, Nathan Rayens, Linh Do, Christine K. Payne, and Scott A. McKinley |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-21025
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Considering experimental frame rates and robust segmentation analysis of piecewise-linear microparticle trajectories Keisha J. Cook \corrauth, Nathan Rayens, Linh Do, Christine K. Payne, and Scott A. McKinley Abstract The movement of intracellular cargo transported by molecular motors is commonly marked by switches between directed motion and stationary pauses. The predominant measure for assessing movement is effective diffusivity, which predicts the mean-squared displacement of particles over long time scales. In this work, we consider an alternative analysis regime that focuses on shorter time scales and relies on automated segmentation of paths. Due to intrinsic uncertainty in changepoint analysis, we highlight the importance of statistical summaries that are robust with respect to the performance of segmentation algorithms. In contrast to effective diffusivity, which averages over multiple behaviors, we emphasize tools that highlight the different motor-cargo states, with an eye toward identifying biophysical mechanisms that determine emergent whole-cell transport properties. By developing a Markov chain model for noisy, continuous, piecewise-linear microparticle movement, and associated mathematical analysis, we provide insight into a common question posed by experimentalists: how does the choice of observational frame rate affect what is inferred about transport properties? \corraddr [email protected]; Tel: +1-864-656-3434. keywords Microparticle, single particle tracking, intracellular transport, changepoint 1 Introduction Intracellular transport has been studied widely to understand the movement and emergent organization of organelles and vesicles within biological cells [ 55 , 64 , 46 , 3 , 20 , 22 , 61 ] . A variety of single particle tracking (SPT) techniques have been developed to capture individual microparticle trajectories as they evolve over time, both in vitro and in vivo [ 5 , 62 ] , using methods that are shaped by a balance between emerging technological capabilities and the biological constraints of particular systems of interest. These methods have produced and will continue to produce fundamental insights into how micro- and nano-scale interactions accrue to produce whole-cell phenomena [ 1 , 34 , 7 , 71 , 32 , 44 , 45 , 26 , 3 , 28 , 65 , 66 ] . For intracellular transport mediated by processive molecular motors, the motion of biomolecular cargo is marked by switches between stationary and motile states. Time spent in either state can be affected by the propensity of motor proteins to either bind or unbind from microtubules, and their ability to process along microtubules when bound. The switches in behavior occur on a time scale of seconds, but transport of individual cargoes across a cell may require minutes, hours, or days. As a result, the temporal resolution at which location observations are made, or simply the frame rate , has a profound impact on how trajectories are perceived. If a motor-cargo complex switches states on the order of once per second, then it is necessary to capture images at least 10 to 20 times per second to resolve the different behaviors. On the other hand, if the frame rate is on the order of once per 10 seconds or more, then switching is not evident and paths can look effectively Brownian. Essential to these investigations are an array of mathematical models that link the transient and rapidly changing micro-scale interactions to stable features at the whole-cell scale. Two of the most popular ways to learn the parameters of these mathematical models are through segmentation analysis and mean-squared displacement (MSD) . Segmentation analysis is typically used for trajectories observed at “fast” frame rates where the hope is that the multitude of motor-cargo (or microparticle) behaviors are highly resolved and can be distinguished reliably [ 25 , 70 , 29 , 8 , 42 , 48 , 11 , 4 , 16 ] . On the other hand, MSD analysis dispenses with the interest in fine-scale behavior and seeks to fit parameters that produce patterns observed at larger time scales [ 56 , 38 , 59 , 60 , 50 , 53 , 40 , 31 , 43 , 42 ] . However, both modes of investigation are prone to different types of systematic error. For MSD analysis, the inferred MSD curves are, by definition, averaging over multiple behaviors. It can happen that multiple fine-scale stochastic models can produce similar emergent MSD profiles. For segmentation analysis, the segmentation step itself is highly vulnerable to bias due to the choice of algorithm, or collection of humans, that do the work. In the work that follows, we focus on segmentation analysis, with an emphasis on identifying the impact of frame rate and the segmentation algorithm on different statistical summaries. Through analysis of simulated and experimental data, we demonstrate a statistical resilience that exists in identifying the proportion of time spent in different states. However, we highlight the limits of the method through a bias-variance tradeoff that emerges across different frame rates. As a consequence we show that, in terms of inference reliability, there is a real difference between two different notions of the term “speed distribution” and identify what appears to be a reliable metric for faithfully characterizing and comparing distinct populations of noisy, continuous, piecewise-linear trajectories. 1.1 Background and experimental context In a typical SPT experiment, the microscope camera captures the positions of a particle over time at a defined number of frames per second (frame rate), which are combined to form a trajectory. Traditionally the frame rate has been decided by the size of target particles, biophysical constraints of the experimental system, the microscope’s focal depth, and limitations on the duration of observation windows [ 59 , 38 , 42 , 62 ] . It is natural to assume that experiments should be run at the fastest possible frame rates, but important tradeoffs exist. The primary tradeoff we will study here arises when there are a limited number of observations that can be made of a single path. Accurate segmentation analysis requires fine enough temporal resolution that state transitions are identifiable, but if trajectories are too short, then too few transitions will be captured and faithful characterization will not be possible. One of the most common techniques in SPT, fluorescence microscopy, is fundamentally limited in this way due to photobleaching [ 19 , 12 , 35 ] . Heuristically, particles of interest are labeled with fluorophores that emit photons when excited by a light source. This light exposure ultimately damages the fluorophores preventing fluorescence. In terms of frame rate, if an experimentalist increases the frequency of observation to resolve finer time scales, the overall lives of fluorophores are shortened, leading to shorter trajectories. Experimentalists sometimes describe this as a ”photon limited” process. A second example of a frame-rate/path-length tradeoff exists in the segmentation algorithms themselves. The complexity of the task in identifying changepoints can grow with the square of the path length, and without proper penalization, substantial numbers of false positives emerge [ 17 ] . Moreover, as more biophysical detail becomes visible, it can become a significant challenge to identify relevant state changes without chasing nuisance phenomena that emerge at the fastest frame rates. This was highlighted in work by Feng et al. [ 18 ] in which the authors tracked individual molecular motor heads using gold nanoparticles at 1000 frames per second, see Fig. 3 in particular. The data is illuminating for studying individual