Identification of fluorescently-barcoded nanoparticles using machine learning.

Area of Science:

• Nanotechnology
• Biotechnology
• Machine Learning

Background:

• Particle barcoding enhances multiplexed bioassays by enabling simultaneous detection of multiple targets.
• Fluorescent barcoding is common for microparticles but challenging for nanoparticles (NPs) due to their small size and heterogeneity.
• Nano-sized carriers are crucial for many bioassays, necessitating effective NP barcoding.

Purpose of the Study:

• To introduce a machine-learning-assisted workflow for writing, reading, and classifying barcoded polymer nanoparticles (PLGA-PEG NPs).
• To address challenges in NP barcoding, including heterogeneity and low signal, for single-particle analysis.
• To provide a framework for optimizing the trade-off between multiplexing capacity and classification accuracy in NP barcoding.

Main Methods:

• Encapsulation of fluorescent markers into PLGA-PEG NPs without altering physicochemical properties ('writing').
• Optimization of confocal imaging techniques for enhanced NP signal detection ('reading').
• Implementation of a machine learning algorithm for classifying barcoded NPs based on fluorescence data ('classification').

Main Results:

• Nanoparticle heterogeneity was identified as a key challenge for barcode separation.
• Förster Resonance Energy Transfer (FRET) from nanoscale confinement effects aided barcode identification.
• The workflow demonstrated successful single-particle level barcoding and classification of NPs.

Conclusions:

• The developed machine-learning-assisted workflow enables robust fluorescent barcoding of nanoparticles.
• Utilizing nanoscale confinement effects like FRET improves barcode accuracy in heterogeneous NP populations.
• This method supports multiplexed bioassays by enabling accurate classification of barcoded NPs.