Lung cancer is the most lethal neoplasia, and an early diagnosis is the best way for improving survival. Symptomatic patients attending Pulmonary Services could be diagnosed with lung cancer earlier if high-risk individuals are promptly separated from healthy individuals and patients with benign respiratory pathologies. We searched for a convenient non-invasive serum test to define which patients should have more immediate clinical tests. Six cancer-associated molecules (HB-EGF, EGF, EGFR, sCD26, VEGF, and Calprotectin) were investigated in this study. Markers were measured in serum by specific ELISAs, in an unselected population that included 72 lung cancer patients of different histological types and 56 control subjects (healthy individuals and patients with benign pulmonary pathologies). Boosted regression and random forests analysis were conducted for the selection of the best candidate biomarkers. A remarkable discriminatory capacity was observed for EGF, sCD26, and especially for Calprotectin, these three molecules constituting a marker panel boasting a sensitivity of 83% and specificity of 87%, resulting in an associated misclassification rate of 15%. Finally, an algorithm derived by logistic regression and a nomogram allowed generating classification scores in terms of the risk of a patient of suffering lung cancer. In conclusion, we propose a non-invasive test to identify patients at high-risk for lung cancer from a non-selected population attending a Pulmonary Service. The efficacy of this three-marker panel must be tested in a larger population for lung cancer.