The exhale breath detection combined bioinformatics analysis method, including transcriptome, metabolic pathway and protein structure, was proposed to identify gas markers for screening and diagnosis of lung cancer. Lung cancer patients and healthy controls' samples were collected to performe GC-MS and ROC curve analysis which obtained ten specific VOCs. Differentially expressed genes were obtained by transcriptome analysis. The differentially expressed genes and relative metabolic pathways were consistent with in vivo biological process, which meant that these VOCs come from the metabolism of lung cancer patient. The sensitivity, specificity and overall accuracy of lung cancer diagnosis model established based on VOCs were 86.2%, 91.2% and 89.6%, respectively. Thus, the proposed method can distinguish normal people and lung cancer patients simply and effectively, providing convenient approach for early screening of lung cancer.