A framework for modeling and inferring tracer diffusion in crowded environments
Jinseok Lee, Tong Lin, Mengyang Gu, Yimin Luo · 2026
Read A framework for modeling and inferring tracer diffusion in crowded environments by Jinseok Lee, Tong Lin, Mengyang Gu, Yimin Luo (2026) 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-2605-04216.txt.
Introduction
Jinseok Lee and three co-authors study how small tracer particles move through crowded settings, such as soft particle suspensions and living cells. As crowding rises, tracers shift from spreading freely to being trapped, and it has been hard to tie that change to the structure around them.
Experiments and a simple simulation agree on how far tracers spread on average. A fast statistical model then predicts that spread from features of the crowd, and the analysis shows the size of open pores matters most. Different overall structures can give the same spread, but pore sizes still track it closely.
Catalog record
| CREATOR | Jinseok Lee, Tong Lin, Mengyang Gu, Yimin Luo |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 81 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.04216
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A framework for modeling and inferring tracer diffusion in crowded environments Jinseok Lee, Tong Lin, Mengyang Gu, Yimin Luo Abstract Tracer diffusion in crowded environments is central to many biological and soft matter systems, but quantitative frameworks for linking tracer motion to environmental structure remain limited. Here, we study the transport of rigid tracers in suspensions of soft particles and within living cells. Experiments reveal a transition from diffusive to confined motion as the matrix area fraction increases. We develop a minimal simulation that incorporates steric exclusion and hydrodynamic hindrance to reproduce the observed mean-squared displacements (MSDs). Using simulation outputs, we train a parallel partial Gaussian process (PPGP) model that rapidly predicts MSDs from matrix geometric variables, including area fraction, particle size, and polydispersity. The PPGP model accelerates predictions by several orders of magnitude relative to simulation and experiments. Analysis reveals that tracer transport is primarily governed by accessible pore sizes and that distinct global structures can produce indistinguishable MSDs. We find that the minimal model can also capture the MSDs of internalized tracer particles in cells. The framework enables rapid inference of structural properties in crowded environments, including transport in the intracellular environment. Introduction The transport of tracers amongst soft particles has vast biological implications [ 32 , 42 ] . For instance, nutrient transport in intracellular environments can be hindered by macromolecules occupying up to 40 vol% of the volume [ 14 , 62 , 12 ] , and a similar degree of hindrance occurs in porous media [ 37 , 48 ] . Tracer trajectories can provide both detailed spatial information and ensemble statistics. The former is useful for quantifying the final distribution of components during transport; the latter, in principle, can help reveal environmental properties using methods such as the generalized Stokes–Einstein relation [ 35 ] . In reality, however, the assumption that the background behaves as a continuum is often not satisfied. New models are therefore needed to describe particle transport in structured environments and to address the inverse problem, i.e. how properties of the environment can be inferred from tracer trajectories and their ensemble statistics. Microrheology is a versatile and powerful technique for probing the rheological properties of complex fluids at microscales [ 15 , 35 ] . Passive microrheology relies on the analysis of the thermal motion of embedded tracers. It has found widespread applications in soft matter physics, biology, and materials science due to its ability to provide local, dynamic information that complements traditional bulk rheological measurements [ 33 ] . One of the most widely used approaches for analyzing particle trajectories is multiple-particle tracking (MPT) [ 13 , 36 ] , which is straightforward to implement but still relies on several user-defined parameters for particle localization and trajectory linking (e.g., radius brightness of particles and the search radius). MPT’s effectiveness diminishes in highly heterogeneous environments, since spatial variability makes it difficult to choose an appropriate, constant cutoff radius for linking trajectories. This becomes a significant concern when precise differentiation between subtle variations within a series of data is essential. On the contrary, Fourier-based techniques, such as differential dynamic microscopy (DDM) [ 9 , 1 , 17 , 31 ] have also been developed to extract dynamic information from microscopy images. By encoding the dynamic information through the intermediate scattering function and coupling with noise models for uncertainty quantification [ 21 ] , DDM can surpass the performance of particle tracking methods in some scenarios, such as optically dense systems. Furthermore, as DDM does not require segmenting particles from background noises and linking their trajectories; therefore, it enables high-throughput processing of experimental series without the need for tuning parameters as in MPT [ 9 ] . More recently, the ab initio uncertainty quantification (AIUQ) framework was proposed in [ 20 ] , which builds a generative probabilistic model and derives the maximum marginal likelihood estimator of the parameters, which optimally integrates information across all wave vectors and removes the need to choose a specific range of wave vectors, as conventionally required in the model fitting by DDM. Particle trajectories have revealed unexpected phenomena that depart from classical Brownian descriptions. In some cases, although the tracers’ mean-squared displacement (MSD) grows linearly in time (Fickian), the displacement probability distribution (also called the Van Hove distribution) deviates markedly from a Gaussian distribution [ 55 , 54 , 24 ] . Correspondingly, the MSD decreases in magnitude compared to pure solvents, which has been attributed to spatial heterogeneities and time-dependent dynamics in the microenvironment [ 24 ] . As the matrix area fraction \Phi increases, the tracer MSD shows a clear transition from diffusive to subdiffusive behavior [ 26 ] , observed in both monodisperse [ 46 , 45 ] and polydisperse [ 11 ] systems. Size differences between the matrix and tracer, and the nature of their interactions dictate the diffusive-to-sub-diffusive transition [ 30 ] . In confinement, the tracer MSDs develop characteristics that are similar to an Ornstein–Uhlenbeck (OU) process [ 52 ] , showing an intrinsic mean-reverting tendency that naturally reflects the excluded volume effect imposed by the matrix particles. Similar MSDs have been reproduced using positional resetting with waiting times drawn from a power-law distribution [ 56 ] . However, experimental confinement is typically less extreme, and the long-time MSD plateau is more gradual compared to the classical OU process or even completely suppressed. In fact, simulations incorporating stochastic hopping probabilities have recovered the observed diffusive MSD without requiring detailed geometric information [ 60 ] . Later studies highlighted the key role of matrix mobility [ 47 , 44 ] , linking non-Gaussian displacements to a transient short-time subdiffusive regime that precedes the long-time Fickian regime by several decades [ 39 ] . Consistently, tracer MSDs often display two diffusive regimes separated by an intermediate subdiffusive phase [ 44 , 49 ] . Despite these advancements, several fundamental questions remain unresolved. Individual tracer trajectories are strongly influenced by local geometry. Beyond the area fraction \Phi , the MSD is also affected by matrix particle size \bar{R} , polydispersity p , and matrix particle mobility. Systematic variations of these variables in simulation studies lead to divergent conclusions of how each variable impacts the MSDs. For instance, some studies have found that, for a given average particle size and area fraction, tracer diffusion is highly sensitive to polydispersity [ 11 ] . In contrast, other work suggests that the polydispersity in matrix particle sizes has little influence on the diffusion regime at small lag time or on the log-exponential coefficients of the MSDs at longer lag time [ 3 ] . These studies provide limited quantitative insight into which variables, \Phi , \bar{R} , or p , dominate the observed subdiffusive MSDs. At what area fraction does the diffusive-to-subdiffusive transition occur in systems with given geometric variables? And perhaps even less explored, how information about the system can be inferred from tracer trajectories? Addressing these challenges is essential for advancing our understanding of soft particulate liquids, which are ubiquitous in industrial, environmental, and biological systems. In this context, the cell cytoplasm is a densely crowded environment, beyond macromolecules [ 14 ] , it is further structured by the actin cortex, which has a mesh