Background and objective: Gliomas are the most common brain tumors usually classified as benign low-grade or aggressive high-grade glioma. One of the promising possibilities of glioma diagnostics and tumor type identification could be based on concentration measurements of glioma secreted proteins in blood. However, several published approaches of quantitative proteomic analysis emphasize limits of one single protein to be used as biomarker of these types of tumors. Simultaneous multi-protein concentrations analysis giving antibody array-based methods suffer from poor measurement accuracy due to technical limitations of imaging systems.
Methods: We applied Principal Component Analysis (PCA) for series of repeated antibody array chemiluminescence images to extract the component representing relative values of protein concentrations, free from zero-mean noise and uneven background illumination - main factors corrupting evaluation result.
Results: The proposed method increased accuracy of protein concentration estimates at least 2-fold. Decision tree classifier applied to the relative concentration values of three proteins TIMP-1, PAI-1 and NCAM-1 estimated by proposed image analysis method effectively distinguished between low-grade glioma, high-grade glioma and healthy control subjects showing validation accuracy of 74.9% with the highest positive predictive value of 81.2% for high grade glioma and 57.1% for low grade glioma cases.
Conclusions: PCA-based image processing could be applied in protein antibody microarray and other multitarget detection/evaluation investigations to increase estimation accuracy.
Keywords: Advanced imaging analysis; Blood serum; Glioma; Protein antibody array.
Copyright © 2021. Published by Elsevier B.V.