Serum proteomic analysis of lung cancer histologic sub-types

Proc Amer Assoc Cancer Res, Volume 47, 2006 4551 Background: Lung cancer remains the leading cause of cancer-related mortality today. Proteomic approaches are powerful tools to search for differentially expressed proteins because they allow the determination of a panel of biomarkers. Protein expression patterns can be obtained by the SELDI system. The present study was designed to evaluate the SELDI system for (i) evaluating serum proteomic profiles in patients with lung cancer, (ii) optimizing clinico-biological classification of lung cancer sub-types, (iii) identifying and characterizing new relevant proteins in this setting. Methods: Analyses were conducted in a population of 64 histologically confirmed lung cancer patients, 10 small cell lung carcinoma (SCLC), 21squamous cell carcinoma (SCC), 33 adenocarcinoma (AD) cases. To improve the discovery and detection of less abundant proteins, an anion exchange fractionation procedure was performed and serum was separated into six fractions. All fractions were analyzed by SELDI technology. The spectra were generated on weak cation exchange and IMAC-Cu arrays for SELDI analysis. The average coefficient of variance was less than 15%. Univariate statistical analysis was performed using Mann-Whitney test for each pairwise comparison. Protein peaks clustering and classification analyses were made using Ciphergen Biomarker Wizard and Biomarker Pattern software, respectively. Construction of the decision tree classification algorithm was performed using cluster and treewise in order to identify specific proteic profiles of the different histological su-types of lung cancer. Results: Thirteen protein peaks were identified as putative histological differenciation markers, 8 for SCLC, 3 for SCC and 2 for AD differenciation. Decision tree classification algorithm was able to discriminate the SCLC and AD groups using one marker, with a 90% sensitivity and a 50% specificity. Using a combination of 3 markers, discrimination of the SCLC and SCC groups achieved a 66% sensitivity and a 83% specificity. Analysis of the SCC and AD histologic sub-groups using hierarchical cluster analysis showed evidences of the existence of different sub-groups in each histologic population (2 in AD group, 3 in SCC group). Conclusion : These results suggest that serum SELDI protein profiling can distinguish lung cancer histological sub-types with a good sensitivity and specificity, and support the possibility of sub-group differenciation of the lung cancer histological types, as previously described using DNA microarray. Confirmatory studies are ongoing.