Open AccessHigh throughput proteomic analysis and machine learning algorithm identifies DUOX2 (dual oxidase 2) as a novel biomarker for response prediction of concurrent chemoradiotherapy for locally advanced rectal cancer.
Author(s) -
Hyebin Lee
Publication year - 2019
Publication title -
journal of global oncology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.002
H-Index - 17
ISSN - 2378-9506
DOI - 10.1200/jgo.2019.5.suppl.71
Subject(s) - biomarker , kegg , medicine , colorectal cancer , chemoradiotherapy , proteomics , biomarker discovery , computational biology , oncology , cancer , cancer research , bioinformatics , gene , gene expression , biology , gene ontology , biochemistry
71 Background: Although many efforts to predict treatment response of concurrent chemoradiotherapy (CCRT) for locally advanced rectal cancer (LARC) have been made, no molecular has proved to be a robust biomarker. Methods: We performed mass spectrometry-based quantitative proteomic analysis of pretreatment Formalin-fixed, Paraffin-embedded (FFPE) biopsy samples of 13 patients with LARC, who were treated with CCRT followed by curative surgery. Based on pathologic report of surgical specimens, we divided thirteen patients as two response groups: complete response (CR) and non-complete response (nCR) groups. Results: A total of 3,637 proteins were identified and 498 proteins were confirmed as expressed at significantly different levels (DEPs; differently expressed proteins) between these two groups. The gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes pathway (KEGG) enrichment analyses were also performed: the result showed that up-regulated DEPs enriched in biological processes (BP) were significantly different between two groups; immune response, cell migration & motility, protein transport in CR group; amide/peptide biosynthetic process, translation, posttranscriptional regulation of gene expression and detoxification in nCR group. To identify the best classifier to evaluate predictive power of signatures, we employed for different machine learning algorithms to classify samples between CR and nCR groups. As a result, we identified the predictive relevance of dual oxidase 2 (DUOX2) as the strongest predictive biomarker. Conclusions: This study identified a new biomarker, DUOX2, applicable to discrimination between CR and nCR after NACRT for LARC. To our knowledge, the present study provides the first identification of a clinical biomarker for response prediction based on in-depth proteomics and machine learning algorithms.