Radiopathomics Predicts Immunotherapy Response in Gastric Cancer

Study Background and Research Question

Gastric cancer (GC) remains a leading cause of cancer mortality worldwide, particularly prevalent in East Asia. Although immune checkpoint inhibitors (ICIs)—notably those targeting PD-1 and PD-L1—have transformed the treatment paradigm for advanced GC, patient responses are heterogeneous. Many do not benefit from immunotherapy, and reliable biomarkers for predicting therapeutic response are lacking. Conventional markers such as combined positive score (CPS), microsatellite instability-high (MSI-H), Epstein-Barr virus (EBV) status, and HER-2 amplification have limited predictive value. Thus, the research community faces an urgent challenge: how can we better predict which GC patients will respond to immunotherapy-based combination therapies?

Key Innovation from the Reference Study

The reference study, published in Cancer Letters, presents a significant advance by developing a multimodal radiopathomics signature (RPS) that combines computed tomography (CT) images and digital H&E-stained pathology images with machine learning. This integrative approach leverages both radiological and histological data to generate a comprehensive, interpretable biomarker. The RPS was rigorously validated across multicenter cohorts, demonstrating superior predictive performance for immunotherapy response compared to conventional biomarkers.

Methods and Experimental Design Insights

The study enrolled 298 patients with advanced GC from multiple medical centers. Baseline data included:

• Pre-treatment CT scans
• Digital pathology slides (H&E-stained tumor sections)
• Clinical and genomic information

Seven machine learning models were trained using radiomic and pathomic features extracted from the images, with a focus on interpretability. The final RPS was constructed by integrating the top predictive features and was evaluated for its ability to predict response to immunotherapy-based combination regimens. Model performance was assessed using area under the receiver-operating-characteristic curve (AUC) in training, internal, and external validation cohorts.

Core Findings and Why They Matter

The RPS achieved high predictive accuracy, with AUCs of 0.978 (training), 0.863 (internal validation), and 0.822 (external validation) according to the reference study. These results surpassed the predictive power of CPS, MSI-H, EBV, and HER-2 biomarkers. Notably, survival analysis (Kaplan-Meier) revealed that the RPS successfully stratified patients into high- and low-risk groups, especially among those with advanced-stage disease or who did not undergo surgery.

Genetic and pathway analyses provided further biological insight. High RPS scores correlated with enhanced immune regulatory pathways and increased infiltration of memory B cells, suggesting that the RPS not only predicts response but also reflects underlying tumor-immune microenvironment dynamics. This supports the growing evidence that integrating data modalities can capture clinically meaningful heterogeneity beyond what single biomarkers provide.

Comparison with Existing Internal Articles

The radiopathomics approach aligns with mechanistic strategies discussed in internal articles such as "PP 1: Enhancing Precision in Immuno-Oncology Signal Dissection", which explores how small-molecule inhibitors like PP 1 can facilitate detailed analysis of oncogenic and immune signaling. Both the reference study and internal discussions emphasize the value of multiparametric data integration for deciphering complex cancer biology. Similarly, "Strategic Disruption of Src Family Kinase Signaling" highlights translational frameworks for targeting signaling networks—such as Src family kinases—underpinning tumor progression and immune modulation. These internal resources underscore the experimental and translational opportunities enabled by combining advanced imaging, computational models, and targeted molecular tools for cancer research.

Limitations and Transferability

Despite robust validation, the RPS’s generalizability may be constrained by cohort composition, imaging protocols, and institutional differences in pathology processing. The study focused on baseline (pre-treatment) data, which may not capture dynamic changes during therapy. Further, while the RPS correlated with immune pathway signatures, causality cannot be inferred without functional validation. Transferability to other tumor types or immunotherapy regimens will require similar multicenter prospective studies. Nevertheless, the approach exemplifies how integrating radiologic, pathologic, and computational analytics can advance precision oncology, particularly where traditional biomarkers fall short.

Research Support Resources

For researchers aiming to dissect the molecular mechanisms underlying immunotherapy response or to experimentally validate radiopathomic signatures, chemical tools targeting key signaling nodes can be highly valuable. For example, PP 1 (Src family tyrosine kinase inhibitor) (SKU A8215) is a potent, selective inhibitor of Src kinases—including Lck and Fyn—with demonstrated efficacy in modulating immune and oncogenic pathways. Its application in in vitro and in vivo models can support studies on the inhibition of Src-family kinases in cancer research, T cell activation modulation, and RET oncogene inhibition. Used in conjunction with advanced imaging and digital pathology workflows, PP 1 provides a mechanistically precise means to probe the molecular context identified by radiopathomics signatures. The compound is available from APExBIO with extensive quality control documentation and is suitable for integration into experimental protocols focused on cancer therapy targeting Src kinases.

Protocol Parameters

• In vitro kinase inhibition: Use PP 1 at concentrations between 5–100 nM to selectively inhibit Src family kinases such as Lck and Fyn in cell-based assays, as recommended by product information.
• Cellular signaling studies: Treat RBL-2H3 or T cell lines with PP 1 (10–50 nM) to assess effects on tyrosine phosphorylation or T cell proliferation; adjust dosing based on experimental endpoints and cell type sensitivity.
• In vivo studies: Administer PP 1 according to established dosing regimens for mouse models of tumor progression or immune modulation, ensuring dosing is tailored to study design and toxicity monitoring.
• Storage and handling: Dissolve in DMSO or ethanol per solubility guidelines; store desiccated at 4°C; do not store solutions long-term.