2264P Proteomics-based phenotypic classification of non-smallcell lung cancer

Current precision medicine strategies for non-small-cell lung cancer (NSCLC) rely on individual mutation analyses, evaluation of the tumor mutational burden or specific marker expression (e.g., PD-L1). However, many patients either do not have a suitable targeted treatment or do not respond to the prescribed therapy. Phenotypic proteomic profiling facilitates the understanding of cancer biology beyond genomics. Furthermore, the more rapid targeted proteomics methods may serve as a tool for patient stratification based on proteomic signatures of the tumors. We used our previously published resource describing the proteome landscape of 141 NSCLC tumors ( Lehtiö et al. 2021 ) to identify biomarkers of the six observed proteomic subtypes. Namely, we applied a machine learning-based algorithm to proteome-wide peptide-level data to select peptide biomarkers for NSCLC subtype classification. We also applied quality control criteria to select peptides suitable for rapid analysis using a targeted proteomics, parallel reaction monitoring (PRM), method. Therewith, we are currently developing a PRM assay for identifying and quantifying the selected biomarkers in clinical samples such as resected tumors and biopsies. The initial cohort of 141 NSCLC tumors will then be re-analyzed to train a model for subtype prediction. We identified 200 peptide biomarkers from 174 proteins representing the six proteomic subtypes of NSCLC. Upon cross validating the classifier, we achieved a high average accuracy of 87%. We then acquired heavy isotope-labeled versions of the peptides for absolute quantification of the biomarkers. Our PRM method analyzing the peptide panel quantifies proteomic biomarkers more accurately and reproducibly than other existing proteomics methods. Furthermore, our classification algorithm converts the quantitative data into categorical NSCLC subtype information. We developed a first-of-its-kind proteomics assay that provides multifactorial characterization of NSCLC tumors covering histology, immune landscape, and oncogenic driver pathways. Our method enables the use of proteomics in a clinical setting and the resultant data may be used for patient stratification.