Prediction of hematogenous metastasis risk from molecular classification in gastric cancer

Hematogenous metastasis (HM) poses a significant challenge in the treatment and outcome of gastric cancer (GC). This study identified and characterized molecular subtypes of GC associated with HM to develop predictive models and enhance precision medicine.

We analyzed bulk-RNA sequencing data from 64 micro-dissected primary GC samples to identify distinct molecular subtypes and validated our findings using a patient-derived xenograft (PDX) model. Signature genes associated with HM were identified using a machine-learning survival model and validated across multiple external cohorts. To evaluate potential therapeutic agents for high-risk HM, we analyzed data from the Cancer Cell Line Encyclopedia.

Two molecular subtypes were identified based on gene expression profiles in GC: stemness and gastric subtypes. The stemness subtype was associated with a significantly increased risk of HM. Validation in the PDX model confirmed accelerated HM in the stemness subtype. The machine survival learning model and differential gene expression analysis identified 10 stemness signature genes and 7 gastric signature genes, which stratified HM risk groups based on a 17-gene signature score. High-risk HM groups exhibited significantly decreased hematogenous metastasis-free survival in three external cohorts.

Stemness subtype gastric cancer is associated with an elevated risk of HM. The identification of molecular characteristics and the development of prediction tools offer practical tools in the clinical field and have a high potential for precision medicine.