Serum Fingerprinting-Based Integrative Dual-Omics Machine Learning for Endometriosis-Associated Ovarian Cancer
Dual-omics, by integrating molecular information from two distinct dimensions, can offer more comprehensive perspective for complex disease. Herein, we developed an efficient functionalized mesoporous nanoparticle-coupled laser desorption/ionization mass spectrometry (fMNPLDI-MS) platform capable of extracting high-quality serum metabolic fingerprints (SMFs) and serum peptide fingerprints (SPFs) from trace serum samples within 50 s. Integrating these SMFs and SPFs markedly improved early screening and subtyping of endometriosis-associated ovarian cancer (EAOC) more than single omics. Specifically, leveraging machine learning on the identified 6 metabolites and 6 peptides, the integration strategy attained an area under the receiver operating characteristic curve (AUC) value of 0.989 and accuracy of 93.1%, compared with 0.964/89.7% for metabolomics alone and 0.960/87.9% for peptidomics alone in distinguishing EAOC from benign controls. Similarly, the dual-omics integration strategy achieved an AUC value of 0.875 and accuracy of 86.7%, surpassing individual metabolomics (AUC: 0.869, accuracy: 83.3%) and peptidomics (AUC: 0.733, accuracy: 76.7%), in subtype classification. This high-throughput fMNPLDI-MS dual-omics platform provides a powerful tool for EAOC screening and subtyping, paving the way toward early detection and precision management of this malignancy.
Journal
Analytical ChemistryInstitutions
Authors
Funding
National Key Research and Development Program of China Grant 2018YFA0507501National Key Research and Development Program of China Grant 2024YFA1307503National Key Research and Development Program of China Grant 2024YFC3405402National Natural Science Foundation of China Grant 21425518National Natural Science Foundation of China Grant 22074019