Analysis of Microarray Data from High Grade Gliomas across Multiple Institutions
S Le, S Sanga, BM Broom, K Aldape, V Cristini, ME Edgerton. MD Anderson Cancer Center, Houston, TX; University of Texas at Austin, Austin, TX
Background: We combined gene expression data from HGG's from M.D. Anderson Cancer Center, Henry Ford Hospital, and University of California, Los Angeles (UCLA) and performed a series of data mining experiments to identify causative gene networks that contribute to aggressive behavior in High Grade Glioimas (HGG).
Design: Data was normalized and divided into class 1 (survival of less than 30 weeks from date of diagnosis) and class 2 (survival of greater than 125 weeks from date of diagnosis). Two-thirds of the data were used as a training set and remaining one-third as a test set, each with approximately 72% class1 and 28% class 2 patients. We used SAM to select for genes and constructed a vote based model to predict class. We performed 2-D hierarchical clustering (HC) to determine which rules clustered with accurate predictions on which patient set. Metacore was used to search for functional relationships and common transcription factors.
Results: We achieve acceptable accuracy (80%) with a set of 117 genes from SAM. The rules defined by SAM genes generate three clusters, one for class 2 (good prognosis) and 2 for class 1 (poor prognosis). Metacore identifies multiple functions and transcription factors associated with class 1 genes, including development, morphogenesis, kinase activity and response to hypoxia.
Conclusions: Rule analysis in combination with pathways databases can be used to study interactions in molecular profiles in HGG, which partition into two poor and one good prognosis molecular subtypes. Genes upregulated in the largest cluster of class 1 and most closely networked share multiple transcription factors, including GGR-alpha, c-fos, Lef1, p-53, NK-Kb, Stat1, Stat3. Stat50, PAX2, and NR2E1.
Category: Pan-genomic/Pan-proteomic Approaches to Diseases
Monday, March 22, 2010 1:00 PM
Poster Session II # 226, Monday Afternoon