2024 World Conference on Lung Cancer (WCLC) - Posters-P3.06F.03 Unbiased Proteomics and Multi-Omics Discovery of a Peripheral Blood-Based Classifier for Early Lung Cancer Detection

P3.06F.03 Unbiased Proteomics and Multi-Omics Discovery of a Peripheral Blood-Based Classifier for Early Lung Cancer Detection

Back to course

Pdf Summary

The study investigates a blood-based multi-omics approach for early detection of lung cancer through a sophisticated classifier. Conducted by PrognomiQ, Inc., in collaboration with Johns Hopkins University, the research involved 2,513 participants from 77 different sites in the United States. The study focused on using a multi-omics platform, which integrates proteomics, RNA sequencing, metabolomics, and targeted immunoassays, to analyze peripheral blood samples. <br /><br />Results indicated the extraction of 113,671 peptides and the detection of numerous proteins, RNA transcripts, RNA introns, and metabolites. A machine learning model using these data achieved a high area under the curve (AUC) of 0.91 initially, improved to 0.96 with the integration of multi-omics data. The model demonstrated strong performance with sensitivity rates of 89% for all stages of lung cancer, and up to 100% in some stages, at a specificity of 89%.<br /><br />The classifier's performance was bolstered by integrating various 'omics types, with each contributing significantly to the results. The study design involved data collection carried out in a uniform and blinded manner to maintain objectivity. From the total participants, 1,623 provided complete datasets, with 798 cases assigned to validation, separated from the training dataset to ensure unbiased evaluation.<br /><br />The multi-omics approach, by leveraging comprehensive molecular data, presents a robust potential for early detection of lung cancer, possibly influencing the treatment approaches and improving prognosis. The findings underscore the advanced predictive power and feasibility of blood-based models in clinical settings, particularly for early-stage lung cancer—highlighting a substantial opportunity for improving survival rates with early intervention. The work was funded by PrognomiQ, Inc., with all authors associated with the company.

Asset Subtitle

Brian Koh

Meta Tag

Speaker Brian Koh

Topic Pathology & Biomarkers

Keywords

multi-omics

lung cancer

early detection

PrognomiQ

Johns Hopkins University

machine learning

blood-based model

proteomics

RNA sequencing

metabolomics