An integrated bioinformatics analysis which computes microarray expression profiling database and predicts sub-cellular location of proteins was exploited to identify lung cancer associated plasma membrane and secreted protein coding genes as lung cancer diagnostic or therapeutic targets. Lung-cancer associated genes were screened from nearly two hundred entries of the NCBI GEO microarray data repository on human normal and cancerous lung tissue samples. Sub-cellular location of proteins was determined both by database mining and computation and they are:searching the entries of Gene Ontology database, searching the sub-cellular localization entry of the LOCATE database, and computing for transmembrane-helices and signal peptides by the TMHMM program and the SignalP program, respectively. The genes which code for plasma membrane and secreted proteins and are differentially expressed in cancer vs. normal tissues, are considered as candidate markers for further investigation. The candidate markers were further examined for their expression in normal tissues by the microarray data set in the GEO database for 32 kinds of normal tissues. The putative lung cancer marker genes identified by the in silico computation were verified with clinical specimens by using quantitative real-time PCR of 50 pairs of lung tumor and adjacent normal tissues. Three secreted protein coding genes (SPP1, COL1A1, COL3A1) and one plasma-membrane protein coding gene (CDH3) were identified as lung tumor associated genes which all have the area under curve (AUC) value in the receiver operating characteristic (ROC) greater than 0.92. The bioinformatics analysis suggests that aforementioned lung cancer associated genes are good candidate markers that they are highly expressed in lung cancer tissues but only have residual expression in most normal tissues. The empirical real time qPCR assay of the genes in19 kinds of normal tissue also confirmed the in silico results. To evaluate the sensitivity and specificity of CDH3 protein as a surface marker, 7 pairs of frozen lung tumor/normal tissues and five different kinds of normal frozen tissues were examined by immunohistochemical staining with CDH3 antibody. The antibody stained only the lung tumor tissues but not the normal tissues. The results were consistent with the results attained for the gene transcripts. The biomarkers identified in this study may have great potential to serve as therapeutic and diagnostic targets.

Citation Information: In: Proc Am Assoc Cancer Res; 2009 Apr 18-22; Denver, CO. Philadelphia (PA): AACR; 2009. Abstract nr 2429.

100th AACR Annual Meeting-- Apr 18-22, 2009; Denver, CO