Related Experiment Video
Updated: Jul 4, 2025
08:20
Using Microarrays to Interrogate Microenvironmental Impact on Cellular Phenotypes in Cancer
Published on: May 21, 2019
5.6K
Machine learning-assisted high-content imaging analysis of 3D MCF7 microtissues for estrogenic effect prediction
Hui Li1,2, Haitham Seada3, Samantha Madnick3
1College of Pharmaceutical Sciences, Center for Drug Safety Evaluation and Research of Zhejiang University, Zhejiang University, 866 Yuhangtang Rd, Hangzhou, 310058, China. hui_li@zju.edu.cn.
Scientific Reports
|February 5, 2024
Summary
This study introduces a novel 3D microtissue imaging approach for endocrine-disrupting chemical (EDC) screening. Machine learning accurately identifies estrogenic EDCs, offering a promising framework for risk assessment.
Area of Science:
- Toxicology
- Biotechnology
- Cell Biology
Background:
- Endocrine-disrupting chemicals (EDCs) present significant environmental and health risks.
- Current high-throughput screening methods for EDCs face challenges with numerous uncharacterized chemicals.
- Three-dimensional (3D) cell culture offers more physiologically relevant models for studying chemical effects.
Purpose of the Study:
- To develop and validate a 3D microtissue imaging and analysis pipeline for endocrine-disrupting chemical (EDC) screening.
- To build a machine learning model for classifying EDC exposure based on quantitative imaging features.
- To establish a framework for assessing estrogenic EDC risk using advanced imaging and computational methods.
Main Methods:
- Formation of 3D microtissues using MCF-7 breast cancer cells.
- Exposure of microtissues to model endocrine-disrupting chemicals: estradiol (E2) and propyl pyrazole triol (PPT).
- Establishment of a 3D imaging and image analysis pipeline to extract quantitative features.
- Development of a machine learning classification model using differential imaging features.
- Creation of deep learning-assisted software for automated lumen formation analysis.
Main Results:
- A robust 3D imaging and analysis pipeline was established for estrogen-exposed microtissues.
- A machine learning model accurately predicted E2 and PPT exposure with high AUC-ROC values (0.9528 and 0.9513).
- Automated software effectively characterized microtissue lumen formation, including number and volume.
- The integrated approach reflected the complexity of estrogen receptor (ER) signaling.
Conclusions:
- The study presents an integrated approach combining image feature profiling and quantitative characterization for EDC risk assessment.
- The developed methods provide a promising conceptual framework for evaluating estrogenic EDCs.
- This approach enhances the understanding of functional ER signaling in complex biological systems.
