Abstract 1202: Expression signature for classification of non-small cell lung cancer.

Abstract

Abstract Background: Approximately 85-90% of lung cancers are non-small cell lung cancers (NSCLCs) consisting of adenocarcinoma (ADC) and squamous cell carcinoma (SCC) and several rare types including undifferentiated large cell carcinomas. Most NSCLCs are diagnosed from small biopsies or cytology materials and, even with the use of immunostains, some cannot be classified further and are referred to as NSCLC - not otherwise specified (NSCLC-NOS). For several clinically relevant reasons and for selection of therapy the correct identification of NSCLC typing is important. Specific Aim: To develop a highly specific and sensitive RNA expression signature as an adjunct test for routine pathological classification. Methods: A microarray dataset of 263 resected NSCLC cases (183 ADC, 80 SCC) obtained from MD Anderson Cancer Center and profiled on the Illumina WG-6 V3 platform was used as the learning set. The Cancer Genome Atlas (TCGA; http://cancergenome.nih.gov/) microarray dataset (105 ADC, 205 SCC), which were profiled on an RNAseq platform, was used as the validation set. Histologic typing on the TCGA specimens was performed by light microscopy according to the WHO classification. Materials for immunostaining were not always available. Results: We developed a 44-gene signature which contained many of the genes encoding proteins routinely used in immunostains for NSCLC typing, including TP63, TTF1, SOX2, and several high molecular weight keratins. For testing, a nearest neighbor approach with Pearson metric was used, which yielded two scores (Pearson correlations, ranging from -1 to +1) for ADC and SCC histologies. Scores above 0.4 were considered positive for the respective tumor type, while values below 0.4 were considered to be null (equivalent to poorly/undifferentiated histology). The learning set had 7% discrepancies for the major tumor type and 1% was scored as null. Initial validation of the TCGA data indicated a relatively high percentage of discrepancies (10% of ADC, 12% of SCC). After discussions with TCGA reference pathologists Drs. Travis and Rekhtman and further evaluation by them a revised pathological classification was generated, which included 25 NSCLC-NOS diagnoses. The discrepancies were reduced to 9 % and 4 % for ADC and SCC respectively. The NSCLC-NOS cases were classified as ADC, SCC or null types. Future plans: We are utilizing the expression signature to develop and validate a quantitative nuclease protection assay (qNPA by High Throughput Genomics) that can be applied to small biopsies and 5-micron sections of archival paraffin embedded formalin fixed material in a 96 well format. Summary and significance: The development and application of a sensitive and specific molecular signature for the classification of NSCLC provides an important adjunct test for routine pathological diagnosis, especially for application to small specimens (which constitute about 70% of current diagnostic lung cancer samples). Citation Format: Luc Girard, Ignacio Wistuba, William D. Travis, Natasha Rekhtman, Milind Suraokar, Carmen Behrens, John D. Minna, Adi F. Gazdar. Expression signature for classification of non-small cell lung cancer. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 1202. doi:10.1158/1538-7445.AM2013-1202