Background: Lung cancer is the leading cause of cancer death in most countries. Although early diagnosis and treatment critically influence prognosis, lung cancers are generally only discovered in the late stages of the disease. Objective: Widely-used screening and diagnostic methods are not suitable for preventive screening, and high-throughput technologies based on serum biomarkers are needed. Methods: We screened 501 serum samples, including 224 lung cancer (LC), 126 disease control (DC), and 151 healthy donor (HC) samples for new serum autoantibodies as biomarkers in the early diagnosis of lung cancer. In phase I, we used HuProt (TM) microarrays to perform preliminary serum antibody screening on 24 LC and 24 HC samples. In phase II, we screened 60 LC, 60 DC, and 60 HC serum samples using focused arrays constructed with 22 of the candidate autoantibody biomarkers screened out in phase I. Results: After data modeling and validation, we selected four potential early LC protein biomarker candidates, IL2RB, CENPB, TP53, and XAGE1A, with individual specificities >90% and sensitivities ranging from 21.2% to 32.2%. These four biomarkers had a specificity of >90% and a sensitivity of >65.5% for early LC when they combined in a panel. Further evaluation of these four biomarker candidates using ELISA assays and 273 serum samples (140 LC, 66 DC, and 67 HC) gave similar results (specificity of >91.7%, sensitivity >61.43%). Conclusion: IL2RB, CENPB, TP53, and XAGE1A combined biomarker panel holds potential for rapid screening and improving the diagnosis of early-stage LC, thus potentially also improving its prognosis.