Fig. 1

Schematic representation of the bioinformatics workflow. We constructed a 29-gene EPPI signature from 2,246 samples including primary NSCLC tumors and normal lung tissues using an integrative genomic approach. The EPPI risk scoring system was developed from discovery set comprising early-stage patients. The prognostic and predictive performance of the EPPI risk score was further validated in multiple independent validation sets comprising 2,071 early-stage NSCLC patients