Introduction: Lung cancer is the leading cause of death in the U.S. Non-small cell lung cancer (NSCLC), the vast majority of lung cancers, has two main subtypes: squamous and non-squamous. These subtypes are difficult to distinguish histologically but demand distinct courses of treatment, and applying the wrong treatment can be fatal. This presents a profound need for accurate classification tools to assist clinicians in their diagnoses. To address the need, we have developed a gene expression-based classifier to distinguish squamous and non-squamous NSCLC by using quantitative Nuclease Protection Assay1(qNPA). This assay does not require prior RNA extraction, and is automated to fit within standard clinical practice, thus can be applied easily in local hospital laboratories.

Methods: First, we identified candidate biomarker genes through microarray analysis of both fresh frozen and formalin fixed paraffin-embedded (FFPE) samples from 134 NSCLC patients, plus in silico analysis of six microarray datasets from GEO. We then refined the set of marker genes and developed a Support Vector Machine (SVM) classifier on a set of 161 FFPE samples. We obtained these samples from a variety of sources to model the heterogeneity of samples under both academic and community settings. Finally, we validated the performance of the classifier on an independent set of 97 FFPE samples.

Results: The performance of the classifier was measured by its call concordance with three pathologists' IHC panel consensus reads. The classifier distinguished squamous and non-squamous NSCLC of the 97 FFPE independent samples with 95% accuracy. Two of the five discordant samples were confirmed as positive in our ALK screening assay; ALK fusions are generally limited to adenocarcinomas. The classifier, combining with an ALK screening assay can provide increased accuracy for NSCLC subtyping.

The classifier's robustness was further demonstrated by diluting the tumor content with normal adjacent tissue, with as little as 20% of the original tumor content in the final sample. All diluted samples were predicted correctly. In addition, the estimated class probabilities did not vary significantly by dilution ratio, indicating that the classifications are robust to low tumor content.

Conclusions: We have developed a robust, high-performance classifier for distinguishing squamous from non-squamous NSCLC lung cancer in FFPE samples without the need for prior RNA extraction using qNPA. The performance of classifier was validated by an independent set of cohort obtained from multiple sources.

Reference:

1.Martel et al, Multiplexed screening assay for mRNA combining nuclease protection with luminescent array detection. Assay and Drug Development Technologies (2002),1(1), 61-71.

Citation Format: Eva Wang, Zhenqiang Lu, Krishna Moddula, Mark Schwartz, Chris Roberts, Mary-Beth Joshi, David Harpole, Vijay Modur. Non-small cell lung cancer histological sub-typing by gene expression analysis from FFPE tissue. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 1877. doi:10.1158/1538-7445.AM2014-1877