Use of gene set analysis (GSA) for molecular classification of responders and nonresponders to FOLFOX therapy in colorectal cancer (CRC).
3628 Background: Folinic acid (FOL) fluorouracil (F) and oxaliplatin (OX) chemotherapy is a commonly used therapy for CRC. Lacking in current literature are clinically relevant classifiers for potential responders. GSA technique is statistical method that detects significance of sets of genes, instead of examining a gene-by-gene basis. The objective of this study was to identify differential functionally annotated gene expression profiles associated with response to FOLFOX therapy in CRC tumors using gene-by-gene and GSA approaches. Methods: Genome wide expression profile data were collected on pre-treatment tumor tissues from patients with unresectable CRC receiving FOLFOX therapy (n = 83, Affymetrix HG U133A, GSE28702, Tsuji et al. BJC 2012). Gene expression was compared between responders (n = 42) and Non-responders (n = 41). GSA was conducted on 3272 curated gene sets from the Molecular Signatures Database (Subramanian, Tamayo et al. 2005, PNAS 102, 15545-15550) annotated by biological pathway, biochemical function and clinical behavior. Significant analysis of Microarray (SAM) and GSA (Tibshirani et al.) was done to identify gene sets associated with FOLFOX response. Results: Differential expressions of 23 genes were significantly associated with response, based on a single gene approach (p-value < 0.05). 13 of these were located on Chromosome 17 (p < 0.001). Among these, the top 5 ranked genes included NPEPPS, MBTD1, CEP44, LTA4H and CPNE4 which are involved in metal ion binding and aminopeptidase activity. GSA revealed only 44 out of 3272 gene sets were significantly associated with response, with a false discovery rate less than 25%. Increased expression of B-lymphocyte differentiation and Ras-signalling-related gene sets was associated with responders while mTOR signaling and hematopoietic stem cell-related genes set were associated with non-responders. Conclusions: Our data showed that differential biological pathways could be identified to predict response to FOLFOX therapy for CRC patients. Analysis may be useful to help define clinically relevant biologic subtypes among patients with metastatic colorectal cancer.