Serum multiomics prediction of prognosis and adverse reactions to concurrent chemoradiotherapy in patients with esophageal cancer.

Abstract

Objective: Concurrent chemoradiotherapy (CCRT) is an important treatment for patients with locally advanced esophageal squamous cell carcinoma (ESCC). There is still a lack of reliable means to predict efficacy, prognosis and hematologic toxicity.

Design: We analyzed 127 serum samples before CCRT and 93 serum samples after CCRT from 127 ESCC patients via metabolomics by GC-MS. Combined with Olink proteomics, we constructed models to predict response and survival through machine learning. Multiple linear regression was used to construct hematologic toxicity prediction models. In combination with the proteomics of ESCC, metabolic changes were studied.

Results: A prediction model for the efficacy to CCRT was established via serum metabolomics and proteomics (Train, CR/nCR = 28/50, AUC = 0.9848, 95% CI = 0.9639-1.0000; Test, CR/nCR = 17/15, AUC = 0.8854, 95% CI = 0.7800-0.9908). A survival prediction model was established (n = 109, C-index = 0.7640, 95% CI = 0.7140-0.8140). Linear models for predicting hematologic toxicity were constructed (n = 111, R > 0.7). L-serine is important for the prognosis of patients with ESCC treated with CCRT, and SHMT2 is a key protein in serine metabolism that affects the efficacy of CCRT.

Conclusion: The combination of serum metabolomics with proteomics can effectively predict the prognosis and hematologic toxicity, which can provide important data for patients to choose treatment methods.