# Soft Independent Modeling of Class Analogies for the Screening of New Psychoactive Substances through UPLC-HRMS/MS

## TL;DR

This study uses advanced analytical methods to classify new psychoactive substances, helping forensic scientists identify unknown drugs more effectively.

## Contribution

The novel use of SIMCA classification models with UPLC-HRMS/MS data improves the identification of new psychoactive substances.

## Key findings

- PCA revealed distinct clusters for NPS classes like benzodiazepines and JWH.
- SIMCA models achieved high classification accuracy, especially at lower collision energy.
- External validation confirmed the models' effectiveness with real seized drug samples.

## Abstract

The proliferation of NPS has become a global issue, due
to their
easy availability and ability to bypass drug screening tests. These
substances are particularly concerning because of their unpredictable
toxicological effects and the analytical challenge in identifying
them. The present study combines advanced analytical strategies based
on UPLC-HRMS with multivariate analysis to identify and classify unknown
NPS. Tandem mass spectrometry (MS/MS) spectra of 159 analytical standards
were acquired, retention times and MS data were preprocessed and organized
in separate matrices to obtain a training set (including 75% of the
analytes) and a test set (with the remaining 25%). Principal component
analysis (PCA) revealed distinct clusters for different NPS classes,
such as benzodiazepines, JWH, and PINACA, while others, like cathinones
and fentanyl analogues, showed greater dispersion. Subsequently, soft
independent modeling of class analogies (SIMCA) classification models
were built. The models were validated, achieving optimal values, and
correctly classifying analytes included in the test set, especially
when considering the data obtained at lower collision energy. External
validation was conducted using three real seized drug samples. This
confirmed the models’ ability to classify data not included
in the training set, as reflected in the positive validation parameters
achieved for each model. Although some misclassifications occurred
due to the limited availability of standards for certain classes,
the SIMCA models proved highly effective in identifying NPS, demonstrating
their value as a reliable tool for supporting forensic investigations.