Analysing paired array-comparative genomic hybridisation and gene expression microarray data: A novel method to infer gene relationships


Session type:

Xin Yi Goh1,2, Richard Newton3, Lorenz Wernisch3, Rebecca Fitzgerald1
1MRC Cancer Cell Unit, Cambridge, UK, 2University of Cambridge, Cambridge, UK, 3MRC Biostatistics Unit, Cambridge, UK


Genome-wide array-comparative genomic hybridisation (aCGH) and gene expression microarray data provide an ideal platform to investigate gene relationships that were previously unknown. We hypothesised that amplified and over-expressed genes (RGs; regulator genes) might lead to over/under-expressions of genes without copy number changes on different chromosomes (TGs; target genes).


Matched aCGH-expression profiles from oesophageal adenocarcinoma samples (n=54)1-2 were used. Posterior probabilities were calculated for partial correlations using a mixture model3 in GeneNet4 to identify RG-TG pairs that satisfy three conditions: 1) Significant aCGH-expression correlations for potential RG’s; 2) TG’s expressions correlate highly with RG’s aCGH status; 3) TG’s aCGH status correlates poorly with RG’s expression changes. Overall probabilities of potential regulatory relationships were calculated by combining probability scores generated from each condition. Panels of four siRNAs were used to test the effects of knockdown of RGs on potential TGs in cell lines with/without amplifications, followed by rescue siRNA assays and reciprocal vector-mediated over-expression of RGs.


The top regulatory gene-target gene (RG-TG) pairs were FGFR2-JAK1 (score=0.7076), FGFR2-NFIA (score=0.6073), ERBB2-IFIT1 (score=0.5978), ERBB2-BST1 (score=0.5487), ERBB2-SLCO1B3 (score=0.5268) and FGFR2-SAMD12 (score=0.5188). Following RG knockdown by siRNAs in cell lines with amplifications of FGFR2/ERBB2, the expression levels of JAK1 (p=0.0030), NFIA (p=0.0017) and BST1 (p<0.0001) were significantly reduced. Interestingly, these effects were not observed in cells without RG amplifications. To conclusively demonstrate the ERBB2-BST1 regulatory relationship, rescue siRNA assays showed protection against siRNAs, whereby expression levels of ERBB2 and BST1 were significantly lower in cells transfected with empty vectors compared to cells transfected with ERBB2-expressing vector prior to siRNA treatments (p<0.05). In contrast, ERBB2 over-expression in cells without ERBB2 amplifications led to increased BST1 expression (p<0.05).


Paired analysis of aCGH-expression data has revealed novel interactions between genes on different chromosomes. This algorithm is currently being tested on a larger dataset consisting aCGH-expression data from various cancer types.