IDDF2024-ABS-0033 Extracellular vesicles-feature based subtypes of colon cancer: a four-gene signature constructed for prognostic and immunotherapy prediction

<h3>Background</h3> The extracellular vesicles (EVs) participate in the progression of colon cancer (CC), distant metastasis, and drug resistance by mediating the transport of substances between tumor cells and the microenvironment cells. They serve as biomarkers for detecting the stage of tumor development. This study constructed a molecular model based on the characteristics of EVs to predict the clinical outcomes and drug sensitivity of CC patients. The EV feature-based subtypes of CC patients <h3>Methods</h3> We collected the gene sets of EV-associated features (EV components, EV biogenesis, EV uptake) from previous research. EV-associated features of each patient of the Cancer Genome Atlas (TCGA) dataset were quantified by Gene Set Variation Analysis (GSVA). The patients were separated into two subtypes by Non-negative Matrix Factorization (NMF). Cox regression and LASSO regression were employed to construct a signature for prognosis and drug sensitivity prediction, which was validated in four independent cohorts. <h3>Results</h3> The expression levels of EV features among different cell populations suggest that different clusters achieved specific physiological processes through distinct EV-related pathways (IDDF2024-ABS-0033 Figure 1. The EV feature-based subtypes of CC patients (A-B)). Therefore, EV feature-based subtypes have been established based on the characteristics of EVs within the CC population. These two subtypes exhibit distinct prognoses, tumor microenvironments, microsatellite instability, tumor mutational burden, and sensitivity to immunotherapy (IDDF2024-ABS-0033 Figure 1. The EV feature-based subtypes of CC patients (C-D)). The four-gene signature was constructed based on the disparity of EV feature-based subtypes (IDDF2024-ABS-0033 Figure 1. The EV feature-based subtypes of CC patients (E)). A nomogram, combining risk score, age, and N stage, was developed, showing strong clinical applicability across four independent cohorts with a time-dependent area under the curve (AUC) (&gt;0.75) (IDDF2024-ABS-0033 Figure 1. The EV feature-based subtypes of CC patients (F-G)). <h3>Conclusions</h3> We proposed a novel EV feature-based subtypes of CC patients and identified two CC subtypes with distinct biological characteristics, various prognoses, and drug responses. We also constructed a four-gene signature showing promising clinical implications for predicting clinical outcomes and drug responses.