Serum-Based Detection of Pancreatic and Ovarian Cancer via a Nanoparticle-Enhanced Fluorescence Array and Machine Learning

TL;DR

This study explores using nanoparticle-enhanced fluorescence and machine learning to detect pancreatic and ovarian cancer from blood samples, showing promising diagnostic accuracy.

Contribution

A novel array-based assay methodology combining nanoparticle-enhanced fluorescence and machine learning for serum-based cancer detection.

Findings

3-NH3+ nanoparticles showed over 80% AUC for pancreatic cancer detection.

2-OH nanoparticles achieved over 70% AUC for ovarian cancer detection.

Boosting algorithms outperformed other methods in classifying cancer states.

Abstract

Background: Early detection of oncological
diseases
such as pancreatic ductal adenocarcinoma (PDAC) and ovarian cancer
(OV) is pivotal for successful treatment but remains a significant
challenge due to the lack of sensitive and specific diagnostic tests.
Fluorescence spectroscopy, enhanced by the interaction of serum proteins
with nanoparticles (NPs) based on linear–dendritic block copolymers,
has emerged as a promising technique for the noninvasive detection
of these malignancies. This study introduces a novel array-based assay
methodology to evaluate the diagnostic capabilities of various NPs
within serum samples using fluorescence. Methods:
We synthesized three types of NPs (1-SH, 2-OH, 3-NH3
+) and analyzed their fluorescence spectra in serum samples
from patients with PDAC, OV, and control subjects. The samples were
excited at 330 and 350 nm wavelengths to obtain their fluorescence
emission spectra. An array of machine learning algorithms was applied,
including boosting and tree-based methods, to assess the ability of
the spectral data to discriminate between pathological and nonpathological
states. The algorithms’ performance was measured by the area
under the receiver operating characteristic curves (AUC). Results: The fluorescence spectra revealed distinct patterns
for PDAC and OV pathologies. 3-NH3
+ NPs exhibited
the highest differential capacity with AUCs exceeding 80% for PDAC
across all algorithms, except one. 2-OH NPs showed a strong discriminatory
ability for OV with AUCs over 70%, utilizing all but one of the algorithms.
1-SH NPs, however, did not significantly increase differentiability.
Boosting algorithms generally outperformed other methods, indicating
their suitability for this diagnostic approach. Conclusions: The proposed assay array methodology enables the systematic evaluation
of NPs’ diagnostic potential using fluorescence spectroscopy.
The differential interactions between NPs and serum proteins specific
to PDAC and OV highlight the method’s capability to discern
pathological states. These findings suggest a path forward for developing
NP-assisted fluorescence spectroscopy as a viable tool for cancer
diagnostics, potentially leading to earlier detection and improved
patient outcomes.

Figures