Microfluidics-based label-free SERS profiling of exosomes with machine learning for osteosarcoma diagnosis

Area of Science:

• Biomedical Engineering
• Nanotechnology
• Oncology

Background:

• Osteosarcoma (OS) diagnosis requires early detection to improve patient survival rates.
• Exosomes are promising noninvasive biomarkers for early cancer diagnosis.
• Current diagnostic methods for OS can be invasive or lack sensitivity.

Purpose of the Study:

• To develop and validate a microfluidic device for purifying and analyzing plasma-derived exosomes for label-free osteosarcoma diagnosis.
• To establish a noninvasive method for differentiating OS exosomes from healthy individuals using surface-enhanced Raman spectroscopy (SERS).
• To construct a machine learning model for accurate OS diagnosis based on exosome SERS profiles.

Main Methods:

• Isolation and purification of exosomes from plasma using a size-dependent microfluidic chip with tangential flow filtration.
• Label-free SERS analysis of exosomes utilizing a nanoarray chip with gold nanoparticle (GNP) self-assembly monolayers.
• Differentiation of OS exosomes based on intrinsic SERS signals.
• Development of a machine learning diagnostic model using exosome SERS data.

Main Results:

• Achieved a high exosome recovery rate of 82% using the microfluidic chip.
• Successfully differentiated exosomes from different OS cell types and between OS patients and healthy donors via label-free SERS.
• Constructed a machine learning model that diagnosed OS with 93% accuracy.
• Demonstrated the potential for label-free, noninvasive OS diagnosis using plasma exosomes.

Conclusions:

• The developed microfluidic SERS-based approach enables noninvasive and precise diagnosis of osteosarcoma.
• This method shows significant potential as a diagnostic tool for osteosarcoma and can be generalized to other diseases.
• Label-free exosome profiling offers a sensitive and specific strategy for early cancer detection.