TopDeck is a novel method for identifying fusion gene samples from The Cancer Genome Atlas (TCGA) using recount3. Developed to create a subset of potential fusion gene samples that can then be run on more accurate fusion finders, TopDeck combines a gene overexpression analysis with a novel exon comparison analysis to identify a list of potential fusion genes.
TopDeck allows for the analysis of a single 3' fusion gene partner, by identifying if it is overexpressed, and if there is a significant increase in exon expression after the expected fusion gene breakpoint. Developed on 10 genes which act as common 3' fusion genes in TCGA (ALK, ARHGAP26, ERG, ETV1, MAML3, NTRK3, RARA, RET, TACC3, TFE3), TopDeck had an overall sensitivity of 36.83%.
TopDeck is the culmination of an Honours thesis for Monash University in collaboration with the Peter MacCallum Cancer Centre.
A demonstration of the gene overexpression method using the 3' gene TACC3 is available in the file overexpression_workflow.Rmd, along with the functions used.
Overexpression was calculated as samples above the 95th percentile, after a preliminary analysis with Tukey's definition of outliers failed to identify a significant amount of fusion gene samples (outliers.Rmd).
To continue to improve the sensitivity of gene overexpression, a preliminary investigation into incorporating the shape of the distribution was conducted, and is summarised in distribution_shape.Rmd. To reduce the false positive rate of the 95th percentile method (identifying 5% of every cancer type as a potential fusion), a preliminary investigation was also conducted into approaches to filter to cancer types of interest through comparison to non-cancer data. This approach is summarised in low_expression.Rmd.
A demonstration of the change in exon expression method using 3' gene TACC3 on bladder urothelial carcinoma samples is available in file tacc3_demonstration.Rmd, with functions used explained.
Two different change in exon expression methods were used, from change in individual exon expression and change in average exon expression.
- The Z-scores for all samples investigated using the change in individual exon expression are available in file: all_diff_objects.Rdata
- The Z-scores for all samples investigated using the change in average exon expression are available in file: prop_cum_all.Rdata
The true positive rail_ids are available in files:
- truepos_diff.csv (individual exon expression)
- truepos_prop_cum.csv (average exon expression)
The false positive rail_ids are available in files:
- falsepos_diff.csv (individual exon expression)
- falsepos_prop_cum.csv (average exon expression)
The change in exon expression method uses functions from exon_expression_functions.R.
A bibliography for R packages used is available in bibliography.R