[1689] A Paradigm for Biomarker Discovery Combining Expression Array Data Mining and Tissue-Based Validation

MD Hofer, S Kageyama, T Tanaka, K Nakagawa, T Hori, LR Chirieac, J Fukuoka. Brigham and Women's Hospital, Boston, MA; Toyama University Hospital, Toyama, Japan

Background: Lots of published expression array data is available on the internet but not often utilized. We describe the in silico analysis of expression data and validation of dysregulated genes on a multi-tumor tissue microarray.
Design: Expression data was mined using the internet-based platform Oncomine for genes dysregulated in cancers that are routinely and experimentally evaluated and had limited reports on PubMed. Dysregulation was validated in 1150 patient samples of 14 tumor types with IHC on a 4-tier system and correlated with clinical (chi-square test) and outcome data (Kaplan-Meier analysis).
Results: Of 65 genes initially, Calgranulin B (mac387), Lamin B, and CD138 met all of our selection criteria. Expression varied in tumor types.

Mean Staining Intensities

	Calgranulin B	Lamin B	CD138
pulmonary SCC (PSCC)	1.1	1.1	1.6
pulmonary adeno ca (ADC)	0.5	1.0	0.7
breast ca	0.2	1.1	0.8
RCC	0.1	1.0	0.4
bile duct ca	0.2	1.0	0.8
thyroid ca	0.0	1.0	0.8
HCC	0.1	0.9	1.6
colon adeno ca (CADC)	0.2	1.2	1.1
gastric ca	0.1	1.1	0.7
prostate ca (PrCA)	0.0	0.5	0.2
pancreatic ca (PaCA)	0.1	1.0	0.4
urothelial ca	0.5	1.5	1.5
ovarian ca (OCA)	0.2	1.0	0.5
endometrial ca (ECA)	0.3	1.2	0.5

Calgranulin B expression was significantly associated with stage in ECA, lymph node metastases (LNM) in OCA, lymphovascular invasion (LVI) in PSCC (all p<0.05) and survival in CADC (log-rank p=0.0261, mean follow up 34 mo). CD138 expression was significantly associated with LVI in PADC, stage in PSCC, stage and LNM in PrCA (all p<0.05) and survival in PADC (log-rank p=0.001, mean follow up: 54 mo) and CADC (log-rank p=0.006). Lamin B expression was significantly associated with stage in PaCA, and stage, LVI, and LMN in PADC (all p<0.05) in which it was also associated with survival (log-rank p=0.003).
Conclusions: We demonstrate the use of internet-based gene expression analysis of published data and validation in tissue samples as approach to facilitate biomarker discovery. Using established and novel IHC markers, we found associations with tumor progression and survival for several tumors. We believe that pathologists are in a unique position to validate previously reported dysregulated genes in patient tissue thus allowing for streamlined biomarker discovery.
Category: Special Category for 2009 - Pan-genomic/Pan-proteomic approaches to Cancer

Tuesday, March 10, 2009 9:30 AM

Poster Session III # 229, Tuesday Morning