Easy Identification of <i>Leishmania (Leishmania) amazonensis</i> and <i>Leishmania (Viannia) braziliensis</i> Species by Using Fourier-Transform Infrared Spectroscopy and Machine Learning Algorithms

Leishmaniasis is a neglected tropical disease requiring accurate species identification to ensure proper clinical management and epidemiological surveillance. Accurate species and strain identification depends on molecular and biochemical tools. While these conventional techniques are effective, they are often costly, time-consuming, and inaccessible in low-resource settings. In this study, we evaluated the potential of Fourier-transform infrared (FTIR) spectroscopy, combined with machine learning algorithms, for the discrimination of <i>Leishmania amazonensis</i> and <i>Leishmania braziliensis</i> species in liquid cultures. FTIR spectra were acquired from 80 culture samples and preprocessed using standard normal variate (SNV) correction and Fast Fourier Transform (FFT) filtering. Principal Component Analysis (PCA) revealed clear species-specific clustering driven by spectral differences in protein, lipid, and nucleic acid vibrational bands. Support Vector Machine (SVM) models were trained using PCA scores, achieving over 90% accuracy in all tested configurations. The best model, using a linear kernel and the first three principal components, reached 100% accuracy, sensitivity, and specificity in external validation. Our findings demonstrate that FTIR spectroscopy, in combination with SVM, offers a rapid, low-cost, and scalable strategy for the screening and classification of <i>Leishmania</i> species, with promising applications for field and clinical diagnostics.