# Advancing Medical Diagnostics: Rapid, Label-Free Detection and Differentiation of Shiga Toxin Variants in Human Serum Using a Cost-Effective PCA-Assisted SERS Platform

## TL;DR

A new low-cost SERS platform can detect Shiga toxins in human serum quickly and accurately, improving early diagnosis of STEC infections.

## Contribution

A first-of-its-kind PCA-assisted SERS platform for label-free detection of Shiga toxin variants in complex biological samples.

## Key findings

- The platform detects Stx1, Stx2a, and cleaved Stx2a in human serum with a detection limit of 0.007 ng/mL.
- Gold nanoparticle-based SERS substrates are cost-effective and highly sensitive in complex biological matrices.
- PCA-based machine learning improves toxin classification accuracy in clinical diagnostics.

## Abstract

Shiga toxins-producing Escherichia coli (STEC) are zoonotic pathogens causing severe diseases such as hemorrhagic
colitis (HC) and hemolytic uremic syndrome (HUS). Infections caused
by STEC represent a public health concern due to the severity of the
possible outcome and acute mortality. The early diagnosis of the infection
is pivotal to driving a correct therapeutic protocol to limit the
severity of the symptoms. The diagnosis is quite cumbersome, requires
specialized approaches, and thus is rarely performed in the hospital,
being managed by the relevant national reference laboratory, delaying
the administration of the appropriate supportive care. In this context,
the demand for affordable diagnostic tests to be carried out at the
bedside is crucial for providing high-value healthcare. In this study,
for the first time to the best of our knowledge, we developed and
optimized a highly sensitive SERS-based platform that can detect and
identify the two main Shiga toxin variants (Stx1 and Stx2a) as well
as the cleaved form of Stx2a in human blood serum at extremely low
concentrations with limits of detection reaching 0.007 ng/mL (0.1
pM). This method uses affordable, sensitive, and very efficient SERS
substrates based on gold nanoparticle films, made with a cost-effective
bottom-up approach, which are much cheaper than those typically found
in the literature. Our results show that the platform works well in
complex biological samples, offering high sensitivity and specificity.
Moreover, integrating machine learning algorithms, such as principal
component analysis (PCA), enables accurate identification of toxin
types, overcoming the limitations of conventional diagnostic methods.
This innovative approach represents a significant step toward accessible,
rapid, and scalable clinical diagnostics, potentially transforming
the early detection and management of STEC-related infections and
preventing life-threatening complications.