Development and Validation of a Prognostic and Predictive 32-Gene Signature for Gastric Cancer

ABSTRACT: Objective: Develop a genomic assay to assess prognosis and guide therapy for gastric cancer patients.Design: This was a retrospective study of multiple, independent cohorts of gastric cancer patients. We performed a pan-cancer analysis of somatic mutations found in 6,681 patients with 19 different cancers using our machine learning algorithm NTriPath to identify a gastric cancer-specific gene signature. We applied this signature to a 567-patient cohort to establish genomic-based molecular subtypes and then used a support vector machine to build a molecular subtype-based risk-scoring model. We evaluated the prognostic capability of the model and investigated whether molecular subtypes predicted response to adjuvant chemotherapy and immune checkpoint blockade. Results: We identified a 32-gene signature that categorized gastric cancer patients into four subtypes that were independently associated with overall survival and predicted response to therapy. The 5-year overall survival of Group 1 patients who received adjuvant 5-fluorouracil and platinum was significantly worse than patients who did not receive therapy (hazard ratio (HR), 6.80 (95% CI, 1.46-31.6), P=0.015). Group 3 patients who received therapy did significantly better than those who did not (HR, 0.28 (95% CI, 0.08-0.96), p=0.043). The subtypes also predicted response to immune checkpoint inhibitors in patients with metastatic or recurrent disease. Groups 1 and 3 patients had significantly higher response rates (44-100%) than patients in Groups 2 and 4 (0 to 13%;P < 0.01).Conclusions: The 32-gene signature is a promising prognostic and predictive biomarker to guide the use of adjuvant chemotherapy and immune checkpoint inhibitors in gastric cancer patients

PROVIDER: EGAS00001005588 | EGA |

REPOSITORIES: EGA