Proteomic analysis of lung cancer tumors

Abstract

7028 Background: In this study we assess the utility of direct tumor tissue MALDI-MS in a large prospective collection of surgically resected lung cancers to distinguish cancer from non-cancer, histology, occult lymph node involvement, and survival. Methods: 175 non-small cell lung cancer specimens and 62 histologically normal lung tissues obtained at the time of surgery were used in this analysis. Twelve micron thick frozen sections were placed on conductive glass slides. Sections were stained with Cresyl Violet and matrix applied to areas identified by a pathologist under microscopic guidance as containing greater than 80% tumor before three separate areas involved with tumor was analyzed by MALDI MS. Relative intensities of selected peaks were used for class comparison. A class prediction model was built based on the weighted flexible compound covariate method of analysis (WFCCM). Results: We created a prediction model from a training cohort consisting 81 tumors and 19 histologically normal tissue samples. A total of 221 peaks were used for statistical analysis. In tumor/normal discrimination, 46 peaks were used for the prediction model (p<0.0001), while, 22 were used (p<0.005) to predict histology, 11 (p<0.05) to predict nodal involvement, and 14 (p<0.005) in survival prediction. Using these prediction models, classification accuracy was 90% in normal/tumor discrimination, 81.8% in histology and 61.7% in nodal involvement prediction. In the survival prediction model, patients with longer than median survival could be distinguished from those with shorter than median survival (p<0.0001, Log-rank test). We then validated the same features in a blind test set from the remaining 93 tumors and 43 normals. For tumor/normal discrimination, prediction accuracy in test cohort was 94.3%. Histology prediction accuracy was 93.9% in predicting squamous cell carcinoma. Nodal involvement prediction accuracy was 49.3%. Longer or shorter median survival was also predicted in this set with statistical significance p<0.08. (Log rank test). Conclusion: We report the analysis of a large set of tumor and normal samples by MALDI MS and confirm similar accuracy in tumor/normal, histology and survival discrimination to previous our report. This project was supported by the Lung SPORE P50 CA90949 (DPC) and NCI 5R33CA86243 (RMC). No significant financial relationships to disclose.

Proteomics analysis in lung cancer: Challenges and opportunities

“…2). 20 These results are showing feasibility of prediction based on MALDI-TOF protein profiling analysis.…”

Review: Adjuvant therapy for resected non-small cell lung cancer

“…Assessment of the proteome profile of the tumor could be a tool for both diagnostic and prognostic purposes. Recently Kikuchi and colleagues presented the results of the proteomic analysis of lung cancer tumors [Kikuchi et al 2006]. The protein profiles of 175 lung cancers and 62 histologically normal lung tissues were analyzed.…”