Application of Surface-enhanced Raman Spectroscopic Method Assisted by Chemometrics in Rapid Assaying of Growth and Testosterone Hormones in Blood

Abstract

This thesis describes work that was done involving the use of surface-enhanced Raman spectroscopy (SERS) with artificial neural network (ANN) in detection and quantification of growth hormone (GH) and testosterone (TE) hormones in blood of non-injected and hormone-injected Sprague Dawley (SD) rats. These hormones are often abused by athletes to boost their performance and win competitions. However, current detection methods possess limitations (such as use of large sample amounts, complex sample preparation, and are time consuming) that are overcome by SERS technique. The SERS substrates employed were based on colloidal silver nanoparticles (AgNPs) synthesized by laser ablation of silver granules in pure water. These substrates were determined to be spherical with a uniform size (34 nm) using a Scanning Electron Microscope (SEM). They were also found to be chemically stable within the first few days of synthesis and resulted in an enhancement by a factor of 10 as compared to Raman intensity of blood without the SERS substrates. The GH and TE hormones concentration-sensitive SERS bands, identified through Principal Component Analysis (PCA) and Analysis of Variance (ANOVA), exhibited significant intensity changes correlating with hormone concentrations in blood. These bands were those centered around 1490 and 1510 cm-1 for GH, 1614 and 1636 cm-1 for TE; and 684 cm-1 for the hormone mixture (GH+TE) in blood. Using PCA scores from simulate samples (blood with known concentrations of each hormone), two ANN models for each hormone were trained and validated with high coefficients of determination (R2) (greater than 87.71%) and low root mean square error (RMSE) values (less than 0.6436 ng/ml) suggesting accuracy of the models. The minimum LOD deduced were 0.3222 ng/ml (for GH) and 0.1851 ng/ml (for TE) while the minimum LOQ determined were 0. 9766 ng/ml (for GH) and 0. 5611 ng/ml (for TE). This revealed that the LOD and LOQ values obtained for TE enanthate were notably in range with those reported by other analytical methods. In addition, detection limits obtained for GH were higher (by 0.28 ng/ml for LOD and 0.9 ng/ml for LOQ) when compared to ELISA, lower for LOD (by 0.2 ng/ml), and higher for LOQ (by 0.2 ng/ml) for isotope dilution mass spectrometry method. Since these detection limits were very low when compared with the range of GH concentrations typically encountered in the rat blood samples, they may still be acceptable. Each of these calibrated models were then used to predict level concentrations of respective hormones in blood from SD rats injected with GH only, TE only, and both GH and TE hormones. This was done to investigate if it was possible to detect exogenous injection of respective hormones in blood. It was v deduced that the hormone levels of injected rats elevated for some time (up to 44 ng/ml) and declined (to 20 ng/ml) later when compared to those of non-injected rats. The same result was observed when ELISA kits were used. Although the two techniques provided similar results, SERS offers additional benefits of being rapid (about two minutes), using simple sample preparation, small sample amounts, and not hormone-specific. This implies that it is possible to detect exogenous injection of sport dopants using SERS spectral data combined with ANN models. To facilitate near real-time display of results, a graphical user interface (GUI) based on SERS data obtained using a portable Raman spectrometer and ANN models was developed in app designer MATLAB. The GUI provided an interaction panel for quantitative analysis of GH and TE concentrations and displayed the results in about two minutes. The study proposes the development of a customized SERS system, where data is rapidly imported into a GUI for hormone level detection, expanding the applications of SERS in sports science, clinical diagnostics, and biomedical research.