AFM ‐Based Multimodal Data Fusion Machine Learning Model for Nanoparticle Endocytosis Recognition in HepG2 Cells

ABSTRACT The lack of sufficient mechanical data and inadequate integration of multimodal information have hindered a comprehensive understanding of nanoparticle–cell interactions in hepatocellular carcinoma (HepG2) treatment screening. This study developed an endocytosis recognition model for silicon dioxide nanoparticles (SiO 2 NPs) in HepG2 cells by integrating multimodal feature fusion with machine learning algorithms. Cell viability was evaluated using the MTT assay for Iron(III) oxide nanoparticles (Fe 2 O 3 NPs), Iron(II, III) oxide nanoparticles (Fe 3 O 4 NPs), and SiO 2 NPs at varying concentrations, sizes, and exposure times. Safe concentration thresholds were determined to be 70, 30, and 50 μg/mL, respectively. Above these levels, cell survival declined in a dose‐dependent manner. The 10 nm Fe 3 O 4 NPs exhibited higher toxicity than their 50 nm counterparts due to their larger surface area. Moreover, mixed‐valence Fe 3 O 4 NPs showed significantly higher toxicity than stable Fe 2 O 3 NPs, highlighting surface reactivity and redox activity as key toxicity mechanisms. After 24 h co‐incubation with FITC‐labeled SiO 2 NPs, fluorescence microscopy confirmed effective endocytosis. Atomic force microscopy (AFM) revealed that SiO 2 NPs slightly increased nuclear Young's modulus, elevated surface roughness, reduced cell height, and enhanced adhesion force at safe dosage levels. Multimodal data (mechanical and morphological) were fused using feature concatenation with four weighting schemes. Endocytosis recognition models for HepG2 cells toward SiO 2 NPs were constructed using Random Forest, SVM, Logistic Regression, and XGBoost. All models achieved over 90% accuracy on the independent validation set. This approach overcomes the limitations of traditional fluorescence‐based methods and provides a novel strategy for nanomedicine safety evaluation and drug screening optimization.

Authors

Institutions

Publication Details

Journal
Microscopy Research and Technique
Published
2026-11-16
DOI
https://doi.org/10.1002/jemt.70198
Primary Topic
Nanoparticles: synthesis and applications
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

AFM ‐Based Multimodal Data Fusion Machine Learning Model for Nanoparticle Endocytosis Recognition in HepG2 Cells

Mei Yin, Xinghua WANG, Zhengxun Song, Ying Wang et al.
1 links
Microscopy Research and Technique
Nanoparticles: synthesis and applications
article

AFM ‐Based Multimodal Data Fusion Machine Learning Model for Nanoparticle Endocytosis Recognition in HepG2 Cells

Mei Yin, Xinghua WANG, Zhengxun Song, Ying Wang, Hongmei Xu, Liguo Tian
article en

Abstract

ABSTRACT The lack of sufficient mechanical data and inadequate integration of multimodal information have hindered a comprehensive understanding of nanoparticle–cell interactions in hepatocellular carcinoma (HepG2) treatment screening. This study developed an endocytosis recognition model for silicon dioxide nanoparticles (SiO 2 NPs) in HepG2 cells by integrating multimodal feature fusion with machine learning algorithms. Cell viability was evaluated using the MTT assay for Iron(III) oxide nanoparticles (Fe 2 O 3 NPs), Iron(II, III) oxide nanoparticles (Fe 3 O 4 NPs), and SiO 2 NPs at varying concentrations, sizes, and exposure times. Safe concentration thresholds were determined to be 70, 30, and 50 μg/mL, respectively. Above these levels, cell survival declined in a dose‐dependent manner. The 10 nm Fe 3 O 4 NPs exhibited higher toxicity than their 50 nm counterparts due to their larger surface area. Moreover, mixed‐valence Fe 3 O 4 NPs showed significantly higher toxicity than stable Fe 2 O 3 NPs, highlighting surface reactivity and redox activity as key toxicity mechanisms. After 24 h co‐incubation with FITC‐labeled SiO 2 NPs, fluorescence microscopy confirmed effective endocytosis. Atomic force microscopy (AFM) revealed that SiO 2 NPs slightly increased nuclear Young's modulus, elevated surface roughness, reduced cell height, and enhanced adhesion force at safe dosage levels. Multimodal data (mechanical and morphological) were fused using feature concatenation with four weighting schemes. Endocytosis recognition models for HepG2 cells toward SiO 2 NPs were constructed using Random Forest, SVM, Logistic Regression, and XGBoost. All models achieved over 90% accuracy on the independent validation set. This approach overcomes the limitations of traditional fluorescence‐based methods and provides a novel strategy for nanomedicine safety evaluation and drug screening optimization.

Microscopy Research and Technique
Changchun University of Science and Technology (CN)
European Commission, Department of Science and Technology of Jilin Province
Good health and well-being
Nanoparticles: synthesis and applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon