Abstract

Fourier-transform mid-infrared (FT-MIR) spectroscopy is a powerful technique that probes intramolecular vibrations of almost any molecule, enabling the acquisition of metabolic fingerprint of cells, tissues and biofluids (e.g. serum, urine and saliva, etc.), in a rapid (in minutes), simple (without or with minimum sample processing), economic (without consumption of reagents), label-free and highly sensitive and specific mode. Due to the flexibility of the technique, there are diverse modes of spectra acquisition, from classical transmission and transflection, to highthroughput measurements using micro-plates in transmission mode, to fiber optic probes coupled to Attenuated Total Reflection (ATR) detection, enabling in situ analysis, throughout micro-spectroscopy, with spatial resolution, enabling detection of residual analytes and imaging at the sub-cellular level. Due to the composition complexity of biological samples, the mid-infrared spectra are usually very difficult to interpret without the application of complex and sophisticated mathematical and statistical analysis routines, such as: spectra pre-processing methods to minimize noise and other non-informative data that compromise subsequent pattern recognition models; deconvolution methods to resolve overlapped spectral bands; methods to decrease data dimension and features extraction; supervised and non-supervised pattern recognition methods as those based on support vector machines and artificial neural networks. The present work reviews the main acquisition modes, pre-processing and multivariate spectral analysis used in FT-MIR spectroscopy, followed by the application of FT-MIR for the diagnosis of a multitude of diseases. FT-MIR spectroscopy constitutes one of the most promising biophysical techniques for analyzing biological samples, and consequently may be used for diseases prognosis, diagnosis and even for personalized treatment.