Background: Inflammation plays an important role in lung cancer development and cancer therapy.To identify potential protein markers for prognosis in non-small cell lung cancer (NSCLC) patients receiving chemotherapy, the association network of lung cancer relevant inflammatory mediators are used to study inflammation system responsive to chemotherapy in cancer.Methods: To prepare the data for our study, sera of healthy non-smokers and patients with NSCLC were collected at the first hospital admission for diagnosis, and again at the second and third admission following one and two course of chemotherapy.According to protein concentrations measured in multiplexed cytokine immunoassays, different associations between inflammatory mediators and NSCLC were investigated by computational analysis: the potential markers (cytokines, inflammatory mediators) related to adenocarcinoma and squamous cell carcinoma were identified through feature selection method (SVM-RFE) with 10 cross-validation ; the decision trees based on those markers were built to give more readable rules associating to relevant cytokines and cancer types ; furthermore, the inferred co-expressed protein interaction network of those protein markers were used to distinguish patients before and after one and two course of chemotherapy.Results: Based on general statistic analysis, several inflammatory mediators were found to be significantly related to NSCLC, and the relation between NSCLC and the combination of inflammatory mediators were further investigated.Two sets of cytokines prefer to identify adenocarcinoma / squamous cell carcinoma samples from pool of patients respectively.The decision trees based on these two kinds of markers can both achieve about 80% accuracy in leave-one-out cross-validation, so that they would be confident to use to diagnose new NSCLC patients.The more interesting and important thing is that, the co-expressed protein interaction network (CEPIN) of cytokines related to adenocarcinoma show dynamic convergent rewiring during chemotherapy.This means the sub-system (protein network) of focusing cytokines in NSCLC tend to be similar to the status of normal persons after multiple courses of chemotherapy, which is supported by the hierarchical clustering of five CEPINs corresponding to different time points before and after chemotherapy.Conclusions: Based on the disease-specific profiles of inflammatory mediators in groups of NSCLC patients, several inflammatory mediators are found to support cancer prognosis, and especially their association network can validate the effects of multi-course chemotherapy .