This study aims to develop a model to predict response to chemotherapy in gastric cancer using RNA splicing information from tumor tissue.

By analyzing genetic patterns and applying machine learning, the study seeks to identify patients who are less likely to benefit from treatment, helping guide clinical decision-making.

This multicenter observational study aims to develop and validate an alternative splicing (AS)-based model to predict response to 5-FU-based adjuvant chemotherapy in stage II/III gastric cancer.

AS events were identified using TCGA SpliceSeq and UCSC Xena data, and selected candidates were quantified by RT-qPCR.

A predictive model was constructed using Elastic Net-based feature selection and XGBoost, and evaluated in independent training and validation cohorts. An integrated model incorporating clinicopathological factors was also developed.

The primary endpoint is treatment response defined by 3-year recurrence-free survival. Patients with recurrence within 3 years are classified as non-responders, and those without recurrence as responders.

This study aims to establish a clinically applicable biomarker for risk stratification and treatment decision support.

Inclusion Criteria:

• Pathologically confirmed stage II or III gastric cancer
• Underwent curative surgical resection
• Received 5-FU-based adjuvant chemotherapy
• Availability of tumor tissue samples for analysis

Exclusion Criteria:

• History of other malignancies
• Inadequate or poor-quality tissue samples (e.g., contamination)

Patients with stage II and III gastric cancer who underwent curative resection and received 5-FU-based adjuvant chemotherapy across multiple institutions.