The team of Professor Guan Jian from the Department of Radiation Oncology of Nanfang Hospital of Southern Medical University has conducted in-depth analysis of the molecular characteristics of nasopharyngeal carcinoma (NPC) based on a nuclear magnetic resonance (NMR) platform, exploring markers for NPC auxiliary diagnosis and radiotherapy efficacy, with fruitful results. Following the publication of a cover article in the renowned international academic journal Advanced Science (IF:15.1/Q1) (Cover Article | Fine Subtyping of Blood Lipoproteins Assists Screening of High-Risk Nasopharyngeal Carcinoma Populations), they published an article entitled “Plasma Metabolic Profiles-Based Prediction of Induction Chemotherapy Efficacy in Nasopharyngeal Carcinoma: Results of a Bidirectional Clinical Trial” in the renowned international academic journal Clinical Cancer Research (IF:11.5/Q1), using plasma NMR metabolomics analysis to discover reliable biomarkers for predicting the efficacy of NPC induction chemotherapy (IC). The paper particularly acknowledged the support provided by ProteinT in nuclear magnetic detection and analysis.
NPC is a highly metastatic and invasive head and neck cancer originating from the nasopharyngeal epithelium, with incidence rates rising sharply to 20-40 cases per 100,000 person-years in Southeast Asia and southern China. As the main treatment for advanced NPC, induction chemotherapy is highly controversial in terms of predicting its efficacy due to the lack of reliable biomarkers. | Research Methods | The samples used in this study were derived from clinical trial NCT05682703, and a final analysis included 166 samples (127 in the discovery cohort and 39 in the validation cohort). Based on NMR, quantitative analysis of 38 plasma metabolites and 112 lipoprotein subfraction indicators was performed on participants’ plasma samples before and after IC treatment, comprehensively analyzing changes in the metabolic profile characteristics of the two cohorts. Based on imaging data, patients were divided into IC-sensitive and IC-tolerant groups; machine learning was used to screen potential biomarkers closely related to IC efficacy in NPC patients and to develop a prediction model for NPC IC efficacy.
Experimental design and research approach
| Research Results | The results showed that the effectiveness of IC varies among individual patients. The 127 patients in the discovery cohort were divided into three groups based on tumor burden reduction ratio (TBRR): the sensitive group (TBRR ≥12.6%), the tolerant group (TBRR < 0%) and the non-responsive group (0≤TBRR <12.6%). Compared with the other two groups, patients in the tolerant group had the fewest differentially abundant substances before and after IC treatment, indicating that changes in the metabolite profile are related to treatment efficacy.
Analysis of differential substances before and after treatment in different IC efficacy groups
The 127 patients in the discovery cohort were divided into 2 groups based on TBRR: the IC-sensitive group (TBRR ≥12.6%) and the IC-insensitive group (TBRR <12.6%). The patients in the discovery cohort were randomly divided into a training cohort and a test cohort at a ratio of 7:3, and four classifiers—support vector machine (SVM), Lasso, random forest (RF) and XGB—were used to analyze and model the training cohort. Scores above the threshold were defined as the high-risk group, and otherwise the low-risk group. The results confirmed that the XGB model performed best, with an AUC value as high as 0.792. The XGB model was then used to validate this marker in the validation cohort, with an AUC value as high as 0.786; the results showed that the TBRR of the high-risk group was significantly lower than that of the low-risk group.
Multiple machine learning methods assist the discovery of markers for predicting IC efficacy
| Research Conclusion |
This study revealed that dysregulation of plasma lipoproteins may affect the efficacy of IC in NPC patients. The efficacy prediction model constructed based on the plasma lipid subfraction profile and metabolite expression profile has good predictive ability and was validated in the validation cohort. This study provides theoretical support for the formulation of NPC treatment strategies and the discovery of potential targets to enhance the effectiveness of IC.
