Plasma proteomic biomarkers for predicting chemoradiation response in locally advanced cervical cancer

A significant proportion of patients with locally advanced cervical cancer (LACC) experience treatment failure following concurrent chemoradiation therapy (CRT). Identifying patients who are unlikely to respond to CRT may inform personalized treatment strategies. We evaluated plasma proteomic profiles of 39 patients with LACC stratified by CRT response: complete response (CR), partial response/stable disease (PR/SD), and progressive disease (PD). The primary objective was to identify protein differences between complete responders (CR) vs. non-responders (PR/SD and PD). Secondary objectives included subclass analyses (CR vs. PR/SD vs. PD). Nano-liquid chromatography-tandem mass spectrometry with data-independent acquisition (nano-LC MS/MS-DIA) was used and verified by parallel reaction monitoring. Differential expression between CRT response groups was assessed using a Bayesian moderated t-test with minimal probabilistic imputation. RNA-seq data from The Cancer Genome Atlas were analyzed using DESeq2, with functional enrichment performed via clusterProfiler and coexpression networks built with NetworkAnalyst. Nested t-test with ridge regression with 2,000 bootstrap iterations was used to evaluate predictive performance, and model discrimination was assessed by AUC using R. We identified 507 protein groups. Of these, 407 were consistently detected in > 50% of samples. The primary analysis identified eight proteins significantly downregulated in non-responders involved in inflammatory response pathways. Secondary subclass analysis demonstrated distinct proteomic dysregulation in patients with PR/SD compared to CR groups, but not in patients with PD, with enrichment of pathways related to complement activation, NF-κB signaling, and lipid metabolism. Targeted verification of secondary findings using parallel reaction monitoring confirmed a panel of 13 proteins that robustly distinguished PR/SD from CR, yielding an optimism-corrected area under the curve of 0.832 (95% confidence interval of bootstrapped models: 0.789–0.957). Additionally, exploratory integration with tumor RNA-seq data from The Cancer Genome Atlas revealed coordinated dysregulation in pathways involving hypoxia inducible-factor 1 signaling, proteoglycan and choline metabolism, and the complement and coagulation cascade. Baseline plasma protein profiles found only modest differences between overall responder and non-responders. Nevertheless, secondary analyses suggested exploratory potential of proteins in predicting PR/SD to CRT.