Infrared Spectroscopy Coupled with Machine Learning Algorithms to Investigate Vascular Dysfunction in Ovariectomy: An Animal Model Study

Abstract

A decrease in female sex hormone levels in the body impairs vascular endothelium functioning, leading to vascular dysfunction associated with certain diseases. Animal models of ovariectomy are commonly used to understand its effects on vascular (dys)function. Fourier-transform infrared (FTIR) spectroscopy is a technique capable of extracting detailed molecular information and, as such, has been applied to various biological analyses. This study evaluated systemic changes in the ovariectomy model using mid-infrared spectroscopy. Thirty-eight serum samples from adult Wistar rats were analyzed and divided into 18 in the control group (SHAM) and 20 in the ovariectomized group (OVX). Bilateral ovariectomy was performed, followed by euthanasia of the rats after 15 days. The spectral collection was performed using the Bruker Alpha II equipment (Bruker, Germany), preprocessed, and analyzed using unsupervised analysis methods [principal component analysis (PCA)] and supervised analysis methods [partial least-squares discriminant analysis (PLS-DA)] (MATLAB 2023). For the PCA model, combinations between principal components (PCs) 1 to 4 were performed. Nevertheless, none of the PC combinations allowed a clear distinction between the OVX and SHAM groups. The PLS-DA model exhibited 66% sensitivity, 80% specificity, a false positive rate of 20%, and a false negative rate of 33%. The F-score was 0.727 and the accuracy was 72.7%. However, the y-permutation test demonstrated that this result was random. These results indicate that there is no significant difference in the systemic profile of rats subjected to ovariectomy surgery for 15 days when analyzed using mid-infrared spectroscopy.