Proteomic analysis of amniotic fluids using analysis of variance-principal component analysis and fuzzy rule-building expert systems applied to matrix-assisted laser desorption/ionization mass spectrometry

Amniotic fluids from 36 women who had premature or at term deliveries were studied using matrix-assisted laser desorption/ionization-mass spectrometry (MALDI-MS) [F. Hillenkamp, M. Karas, R.C. Beavis, B.T. Chait, Matrix-assisted laser desorption ionization mass-spectrometry of biopolymers, Analytical Chemistry 63 (1991) A1193–A1202. [1]]. The effects of experimental factors of patient, uterine infection, sample preparation, sample handling, three different days and four different studies, were examined using analysis of variance-principal component analysis (ANOVA-PCA) [P.B. Harrington, N.E. Vieira, J. Espinoza, J.K. Nien, R. Romero, A.L. Yergey, Analysis of variance-principal component analysis: a soft tool for proteomic discovery, Analytica Chimica Acta (in press)]. Hotelling T2 95% confidence intervals were introduced as a graphical tool for evaluating the statistical significance of the levels for each factor. ANOVA-PCA gave biomarkers that distinguished amniotic fluids with evidence of chorioamniotic infection from the other amniotic fluids. All women with intrauterine infection, except for one were identified from the MALDI-MS spectra of their amniotic fluids. After querying the clinic, it was discovered that the amniotic fluid of the misclassified patient was incorrectly identified as infected. The biomarkers obtained from ANOVA-PCA corresponded to those obtained from a fuzzy rule-building system (FuRES) [P.B. Harrington, Fuzzy multivariate rule-building expert systems—minimal neural networks. Journal of Chemometrics 5 (1991) 467–486]. The FuRES model was evaluated using 10 Latin-partitions that yielded a sensitivity of 92% and specificity of 100% for detecting patients with uterine inflammation from MALDI-MS measurements of the amniotic fluids. The specificity value contains a patient that was clinically misidentified as infected that was disclosed by querying the FuRES results. An ANOVA-PCA indicated that the solid phase extraction improved the precision of the analysis, but FuRES analysis showed that this procedure resulted in a lower prediction rate. Using the MALDI-MS data obtained from the amniotic fluids of patients who delivered at term or preterm without evidence of infection in the patients, could not be classified by ANOVA-PCA or predicted by FuRES into these groups.