Deep Learning-Decoded Raman Spectroscopy for Hour-Scale iPSC Pluripotency Assessment via Lipid-Protein Biomarkers.

Area of Science:

• Biotechnology
• Stem Cell Biology
• Computational Biology

Background:

• Assessing induced pluripotent stem cell (iPSC) pluripotency is crucial for regenerative medicine.
• Traditional methods are invasive, require labeling, and are time-consuming, limiting dynamic monitoring.

Purpose of the Study:

• To develop a rapid, label-free method for assessing iPSC pluripotency using deep learning and Raman spectroscopy.
• To establish a new quality control framework for iPSCs suitable for clinical applications.

Main Methods:

• Induced pluripotent stem cells (iPSCs) were cultured with medium alterations to modulate pluripotency.
• Raman spectroscopy was used to acquire spectral data from iPSCs at different pluripotent states.
• A one-dimensional convolutional neural network (1D-CNN) was trained on spectral data for pluripotency classification.

Main Results:

• The 1D-CNN model achieved 97.10% average accuracy in classifying pluripotent states.
• Pluripotency anomalies were detected with 98.00% accuracy at 1 hour, before morphological changes.
• Lipids and proteins were identified as key Raman spectral biomarkers for pluripotency.

Conclusions:

• This integrated approach offers a paradigm shift for noninvasive stem cell diagnostics.
• The method provides a fast, accurate, and label-free quality control for iPSCs.
• The findings link iPSC pluripotency to PI3K/AKT pathway activity.