32P Improvement of diagnostic accuracy of breast cancer using multi-protein signature markers through machine learning

We have developed a 3-protein signature blood marker (Mastocheck®) for early diagnosis of breast cancer. The purpose this study is to improve the performance previously markers. Blood from 196 cancer patients and healthy control groups were prospectively collected.Through development biomarker detectable library, PepQuant, peptides that are optimal MS/MS detection selected. After chemically synthesizing these selected proteins, quantified by multiple reaction monitoring (MRM) methods. Seven final proteins derived applying PepQuant library discovery verification. Machine learning algorithms was trained as protein candidates identified between controls. As in previous studies, evaluation conducted based on sensitivity, specificity, accuracy, false positive rate, negative predictive value, value diagnosis. accuracy Mastocheck®, 69.4%, 83.7%, 76.5%, respectively, which relatively similar studies’ results. rate(FPR) rate(FNR) 16.3% 30.7%, value(PPV) value(NPV) 81.0% 73.2%, respectively. During when 7-protein combined with an artificial intelligence(AI) technique analysis, 88.3%, 83.2%, 85.7%, showing superior compared Mastocheck®. FPR FNR 16.8% 11.7%, indicating improved 20% Mastecheck®. In addition, it recognized PPV NPV also 84.0% 87.6%. Through collection new prospective samples, confirmed diagnostic Mastocheck® repeatedly maintained. using signatures AI model showed can be remarkably improved.