High throughput screening of mesenchymal stem cell lines using deep learning.

Area of Science:

• Regenerative Medicine
• Biotechnology
• Computational Biology

Background:

• Mesenchymal stem cells (MSCs) are vital for regenerative therapies but suffer from functional heterogeneity and inconsistent clinical outcomes due to inadequate quality control (QC).
• Current QC methods for MSCs are limited, necessitating advanced approaches for reliable functional screening.
• Machine learning (ML) and single-cell morphological profiling offer potential but require extensive imaging and lack generalizability.

Purpose of the Study:

• To develop and validate an end-to-end deep learning (DL) framework for high-throughput functional screening of MSC lines.
• To address the limitations of existing ML-based methods by utilizing live-cell microscopic images of MSC populations.
• To establish an effective QC strategy for MSCs in clinical biomanufacturing.

Main Methods:

• An end-to-end deep learning (DL) framework was developed using live-cell microscopic images of MSC populations.
• Various convolutional neural network (CNN) models were quantitatively evaluated for classification accuracy.
• MSC lines were classified based on multilineage differentiating stress-enduring (MUSE) cell markers using immunofluorescence staining and FACS analysis.

Main Results:

• The optimized DenseNet121 model achieved high performance metrics: AUC 0.975, accuracy 0.922, F1 0.922, sensitivity 0.905, specificity 0.942.
• The DL framework accurately classified MSC lines into high/low MUSE cell marker groups from multiple donors.
• A total of 6,120 cell images from 8 MSC lines were analyzed, demonstrating the method's robustness.

Conclusions:

• The proposed DL-based framework provides a convenient and high-throughput method for functional screening of MSC lines.
• This approach can serve as an effective quality control (QC) strategy for future clinical biomanufacturing processes.
• The study highlights the potential of deep learning in standardizing MSC therapies and improving clinical outcomes.