Machine learning framework for investigating nano- and micro-scale particle diffusion in colonic mucus
Open Access
- 22 August 2025
- journal article
- research article
- Published by Springer Nature in Journal of Nanobiotechnology
- Vol. 23 (1) , 1-18
Abstract
Biosimilar artificial mucus models that mimic native mucus facilitate efficient, lab-based drug diffusion studies, addressing the costly and challenging preclinical phase of drug development, especially for nano- and micro-scale particle-based colonic drug delivery. This study presents a machine-learning-driven framework that integrates microrheological features into diffusional fingerprinting to characterize nano- and micro-scale particle diffusion patterns in mucus and assess the effect of mucus microrheology on such movements. We investigated the diffusion of fluorescent-labeled polystyrene particles in native pig mucus and two artificial mucus models. Particles (100, 200, and 1000 nm in diameter) with carboxylate- or amine-modified surfaces were tracked during passive diffusion. From each particle trajectory, 20 features —including microrheology-based parameters— were extracted. Based on these features, seven supervised machine learning models were applied to classify or identify similarities among mucus hydrogels. Of these, gradient boosting achieved the highest accuracy. SHapley Additive exPlanations analysis identified creep compliance as the most influential feature in distinguishing the mucus models. In native mucus, smaller negatively charged nanoparticles exhibited the highest mobility, with fewer particles being in the immobile and subdiffusive states. Microrheology data further indicated that larger particles experienced greater restriction owing to the elastic properties of native mucus. In contrast, smaller particles interacted more with the viscous liquid phase. A comprehensive feature-wide analysis revealed that hydroxyethyl cellulose (HEC)-based artificial mucus more closely resembled native pig mucus than the polyacrylic acid-based model. In conclusion, the machine-learning-driven fingerprinting approach, incorporating microrheological features, successfully differentiated the microstructural characteristics and rheological properties of the three mucus models. It also supported the selection of HEC-based artificial mucus as a viable substitute for native colonic mucus. Graphical abstractKeywords
Funding Information
- Uppsala University
This publication has 38 references indexed in Scilit:
- Inferring pointwise diffusion properties of single trajectories with deep learningBiophysical Journal, 2023
- Software Metrics and tree-based machine learning algorithms for distinguishing and detecting similar structure design patternsSN Applied Sciences, 2019
- The role of mucus on drug transport and its potential to affect therapeutic outcomesAdvanced Drug Delivery Reviews, 2018
- Mucus models to evaluate nanomedicines for diffusionDrug Discovery Today, 2014
- Human colonic mucus is a reservoir for antimicrobial peptidesJournal of Crohn's and Colitis, 2013
- Spatial Configuration and Composition of Charge Modulates Transport into a Mucin Hydrogel BarrierBiophysical Journal, 2013
- The Multiple Logistic Regression Model2013
- Encyclopedia of Distances2009
- Micro- and macrorheology of mucusAdvanced Drug Delivery Reviews, 2009
- Real-time multiple-particle tracking: applications to drug and gene deliveryAdvanced Drug Delivery Reviews, 2005