# Plasma immune proteome-based risk score predicts survival in advanced gastric cancer treated with PD-1 inhibitors and chemotherapy

## TL;DR

A blood-based protein test can predict survival in advanced stomach cancer patients treated with PD-1 inhibitors and chemotherapy.

## Contribution

A five-protein plasma immune proteome risk score is developed to predict survival in gastric cancer patients undergoing chemoimmunotherapy.

## Key findings

- A five-protein panel (LTB4R, GBP2, HLA-G, CYBB, HLA-B) was identified as a prognostic marker for survival.
- The risk score outperformed traditional clinical and PD-L1 biomarkers in predicting overall survival.
- A nomogram integrating the proteomic score with clinical variables provided accurate survival estimates.

## Abstract

Blood-based biomarkers that capture systemic immunity could complement tissue-based assays for prognostication in advanced gastric cancer receiving programmed cell death protein 1 (PD-1)–based chemoimmunotherapy. We evaluated whether baseline plasma immune proteomics can stratify clinical outcomes and be operationalized into a clinically usable model.

In a prospective cohort (n=40) treated with first-line PD-1 inhibitor plus chemotherapy, nano–ultra-high-performance liquid chromatography (nano-UHPLC) coupled with Orbitrap data-independent acquisition liquid chromatography–tandem mass spectrometry (DIA LC-MS/MS) was used to profile baseline plasma. Quality control (QC)-filtered protein intensities were median-normalized, log2-transformed, and batch-adjusted as needed. Group structure was assessed by principal component analysis (PCA). Differential expression (two-sided testing; Benjamini-Hochberg false discovery rate [FDR] correction) and functional enrichment were performed, with an immune focus defined using Immunology Database and Analysis Portal (ImmPort) sets. Prognostic screening used univariate Cox proportional hazards regression; features were reduced by least absolute shrinkage and selection operator (LASSO)-Cox and entered into multivariable models. A risk score (linear predictor of z-scaled abundances) was evaluated by Kaplan-Meier analysis and time-dependent receiver operating characteristic (ROC) analysis. A prognostic nomogram integrating the proteomic score with clinical variables was calibrated by bootstrap resampling.

PCA showed outcome-associated separation. Differential testing identified 322 proteins (179 up, 143 down in long-term survivors), including 36 immune-related differentially expressed proteins (DEPs). Penalized modeling selected a five-protein prognostic panel—LTB4R, GBP2, HLA-G, CYBB, HLA-B. The risk score, dichotomized at the cohort median, stratified overall survival (OS) and progression-free survival (PFS) with clear separation. Time-dependent ROC area under the curve (AUC) values for OS at 6/12/18/24 months were 0.850/0.838/0.911/0.844, exceeding age, sex, grade, and programmed death-ligand 1 (PD-L1) combined positive score (CPS). In multivariable Cox models adjusting for clinical covariates, the score remained independently associated with OS. A nomogram combining the score with clinicopathologic factors yielded individualized 6-, 12-, and 18-month OS estimates with good calibration. Median PFS and OS for the overall cohort were 5.5 and 10.0 months, respectively.

Baseline plasma immune proteomics supports a compact, interpretable five-protein risk score that augments clinicopathologic variables for prognostic stratification under PD-1–based chemoimmunotherapy. The model is amenable to targeted assay translation and prospective validation for clinical deployment.