Integrating Multi-Omics Data for Enhanced Prognosis Prediction in Gastric Cancer Post-Neoadjuvant Therapy

Study Protocol: Integrating Multi-Omics Data for Prognosis Prediction in Gastric Cancer Post-Neoadjuvant Therapy

Objective:

To develop and validate an integrative prognostic nomogram for patients with locally advanced gastric cancer (LAGC) undergoing neoadjuvant therapy, combining deep learning-derived radiomic features (DeepScore), transcriptome-based immune scores (ImmuneScore), and ypTNM staging.

Study Design:

A retrospective, single-center cohort study.

Participants:

A total of 179 LAGC patients who received neoadjuvant therapy followed by radical gastrectomy at Fujian Medical University Union Hospital between January 2019 and December 2022. Patients were divided into a training cohort (n = 125) and an independent validation cohort (n = 54).

Data Collection:

Baseline contrast-enhanced CT scans prior to neoadjuvant therapy were used for radiomic analysis. Postoperative tumor RNA sequencing data were used for immune profiling. Clinical and pathological data, including ypTNM stage, were collected from medical records.

Methods:

DeepScore: Extracted from CT images using a ResNet18-based deep learning model. Significant features were selected via univariate Cox and LASSO regression.

ImmuneScore: Calculated from RNA-seq data using the ESTIMATE algorithm to assess tumor immune infiltration.

Nomogram Construction: A multi-omics nomogram was developed using multivariate Cox regression incorporating DeepScore, ImmuneScore, and ypTNM stage.

Validation: Model performance was evaluated using time-dependent ROC analysis (AUC) and Kaplan-Meier survival analysis with log-rank tests in both cohorts.

Primary Outcomes:

Disease-free survival (DFS) and overall survival (OS).

Statistical Analysis:

Survival analyses were performed using Kaplan-Meier and Cox regression models. AUC values were computed for 1-, 2-, and 3-year DFS predictions. All analyses were conducted in R (v4.4.3).