Journal of IiME Volume 1 Issue 2 www.investinme.org Visible and near-infrared (Vis-NIR) spectroscopy: Introduction and Perspectives for Diagnosis of CFS (continued) Table 3. Reprentative chemometrics software and statistical analysis used in Vis-NIR spectroscopy studies MLR: Mutilinear regression PCR: Principal component regression PLS: PCA: Partial least squares regression Principal component analysis KNN: K-nearest neighbors SIMCA: Soft independent modeling of class analogy HCA: Hierarchical cluster analysis Class least squares CLS: consuming, is not required. Furthermore, cross validation steps are also included, and these reduce the overall handling and risk of error during analysis. Application of Vis-NIR spectroscopy for CFS research Several biochemical changes are reported in CFS patients, but there is no clear consensus for any of them. Therefore, the diagnosis of CFS is currently based on clinical symptoms. As this approach relies on experience and skill, CFS can be diagnosed only by limited numbers of medical doctors. To overcome these problems, an additional method using instrumentation to achieve an objective diagnosis is needed. We reasoned that Vis-NIR spectroscopy might provide new insights if patients could be compared with individuals without the disorder. Here, we describe the results obtained when sera from CFS patients as well as healthy volunteers were subjected to Vis-NIR spectroscopy [86]. At the Medical Hospital of Osaka City University, serum samples from 77 CFS patients (33.0 ± 8.8 years old; Male/Female: 29/48), diagnosed on the basis of clinical criteria proposed by CDC were examined [1]. Samples from 71healthy volunteers (41.7 ± 10.4 years old; Male/Female: 33/38) were also used. The sera of the 77 CFS patients and 71 healthy volunteers served as test samples to develop calibration models for PCA and SIMCA. Another 99 determinations [54 in the healthy group (35.9 ± 9.1 years old; Male/Female: 11/7) and 45 in CFS patients (34.9 ± 7.0 years old; Male/Female: 8/7)] were masked and used for predictions. All samples were diluted 10-fold with phosphate-buffered saline and adjusted to a constant volume (2 ml) in a polystyrene cuvette before the Vis-NIR spectroscopic measurements. Three consecutive Vis-NIR spectra were measured at a resolution of 2 nm with an NIRGUN (Japan Fantec Research Institute, Shizuoka, Japan) at 37°C. The spectral data were collected as absorbance values [log(1/T)], where T= transmittance in the wavelength range from 600 to 1,100 nm. Pirouette software (ver. 3.11; Invest in ME Charity Nr 1114035 Infometrics) was employed for all data processing. To minimize differences between spectra caused by baseline shifts and noise, prior to calibration, spectral data were mean-centered and transformed by SNV [84] and smoothing based on the Savitsky-Golay algorithm [85]. To identify the predominant absorbance peaks in the spectra, PCA and SIMCA methods were further applied to develop PCA and SIMCA models for CFS diagnosis. A clear difference in the sera of CFS patients from those of healthy donors was seen in PCA scores using the first principal component (PC1) and second principal component (PC2) (Fig. 3A, B). The SIMCA model allowed correct separation of the Vis-NIR spectra of 209 of 213 (98.1%) healthy volunteers and 220 of 231 (95.2%) CFS patients. SIMCA using Coomans plots demonstrated that classes of sera from the volunteers and patients did not share multivariate space, providing validation for the separation (Fig. 4A, C). Furthermore, masked samples were subjected to Vis-NIR spectroscopy, and predictions made with the PCA and SIMCA models. PCA clearly distinguished the masked samples of the healthy volunteers from those of the CFS patients (Fig. 3C). SIMCA correctly predicted 54 of 54 (100%) volunteers and 42 of 45 (93.3%) patients (Fig. 4B, D). These results suggest that combining Vis-NIR spectroscopy with chemometrics is a promising way to objectively diagnose CFS. They also suggest that an unknown factor or factors present in the serum of all CFS patients could provide important clues as to the agent causing this debilitating disease. We concede that statistically, the results are not robust enough for clinical use at this time. The PCA and SIMCA model was developed from Vis-NIR (continued on page 13) Page 12/72
13 Publizr Home