METHODS: Biochemical analyses, ELISA (enzyme-linked immunosorbent assays), and multiplex assays were conducted for blood sera from patients with RA (n = 32), patients with PsA (n = 28), and the control group (n = 18). ATR-FTIR spectra were collected for lyophilized sera.
RESULTS: The combination of six biochemical parameters (WBC, ESR, RF, CRP, HCC-4/CCL16, and HMGB1/HMGB) allowed the development of the partial least squares discriminant analysis (PLS-DA) model with an overall accuracy (OA) of 80% for test samples. The best separation between RA, PsA, and the control group was obtained utilizing spectral data. Using the interval PLS algorithm (iPLS) specific spectral ranges were selected and a classifier characterized by OA value for test set equal to 88% was obtained. This parameter, for the hybrid PLS-DA model constructed using selected biochemical parameters and a significantly reduced number of spectral variables, reached the level of 84%.
CONCLUSIONS: PLS-DA models developed on the basis of spectral data enable effective differentiation of patients with RA, patients with PsA, and healthy subjects. They appeared to be insensitive to existing inflammation processes which opens interesting perspectives for new diagnostic tests and algorithms for identification of patients with RA and PsA.
方法:生化分析,ELISA(酶联免疫吸附测定),并对RA患者的血清进行了多重检测(n=32),PsA患者(n=28),对照组(n=18)。收集冻干血清的ATR-FTIR光谱。
结果:六个生化参数的组合(WBC,ESR,射频,CRP,HCC-4/CCL16和HMGB1/HMGB)允许开发偏最小二乘判别分析(PLS-DA)模型,测试样品的总体准确度(OA)为80%。RA之间最好的分离,PsA,对照组是利用光谱数据获得的。使用间隔PLS算法(iPLS),选择特定的光谱范围,并获得以测试集的OA值等于88%为特征的分类器。此参数,对于使用选定的生化参数和显着减少数量的光谱变量构建的混合PLS-DA模型,达到84%的水平。
结论:基于光谱数据开发的PLS-DA模型能够有效区分RA患者,PsA患者,和健康的受试者。他们似乎对现有的炎症过程不敏感,这为新的诊断测试和识别RA和PsA患者的算法开辟了有趣的视角。